This Week in Open Source #13

by Daryl Ducharme, Google Open Source

This Week in Open Source for January 23, 2026

A look around the world of open source

Can you believe we're already wrapping up the first month of the year? January is coming to a close. The open source ecosystem is buzzing with activity, from the upcoming community gatherings at FOSDEM in Brussels to new conversations around AI standards and cloud flexibility.

Google Open Source believes that "a community is a garden, not a building". It requires constant tending to thrive. This week, we're looking at how we can all contribute to that growth—whether it's by securing the software supply chain, standardizing AI agents, or simply learning from the legends of our field like Linus Torvalds.

Dive in to see what's happening this week in open source!

Upcoming Events

• January 29: CHAOSScon Europe 2026 is co-located with FOSDEM in Brussels, Belgium. This conference revolves around discussing open source project health, CHAOSS updates, use cases, and hands-on workshops for developers, community managers, project managers, and anyone interested in measuring open source project health. It also shares insights from the CHAOSS context working groups including OSPOs, University Open Source, and Open Source in Science and Research.
• January 31 - February 1: FOSDEM 2026 is happening at the Université Libre de Bruxelles in Brussels, Belgium. It is a free event for software developers to meet, share ideas and collaborate. Every year, thousands of developers of free and open source software from all over the world gather at the event in Brussels.
• February 24 - 25: The Linux Foundation Member Summit is happening in Napa, California. It is the annual gathering for Linux Foundation members that fosters collaboration, innovation, and partnerships among the leading projects and organizations working to drive digital transformation with open source technologies.
• March 5 - 8: SCALE 23x is happening in Pasadena, California. It is North America's largest community-run open source conference and includes four days of sessions, workshops, and community activities focused on open source, security, DevOps, cloud native, and more.
• March 9 - 10: FOSSASIA Summit 2026 is happening in Bangkok, Thailand. It will be a two-day hybrid event that showcases the latest in open technologies, fostering collaboration across enterprises, developers, educators, and communities.

Open Source Reads and Links

• [Article] The state of trusted open source - This review of the state of trusted open source report goes over many statistics. One of the interesting ones is that vulnerabilities most often hide in the smaller dependencies of the larger projects we might be focused on. What does this mean for your approach to security? How should various open source communities deal with this?
• [Blog] Software Heritage Archive recognized as a digital public good - As the Software Heritage Archive celebrates its 10th anniversary, the Archive has scaled to protect over 27 billion unique source files, even solving the "2PB problem" by deploying protocols that compressed 78TB of graph data into a 3TB research dataset. This ensures that humanity's executable history remains a global commons rather than a proprietary secret, aligning with our belief at Google that Code is for today, Open Source is forever.
• [Blog] Agent Definition Language: The open standard AI agents have been missing - The Agent Definition Language (ADL) creates a clear, shared way to describe AI agents so they work well across different systems. This helps teams understand what agents do, how they behave, and how to govern them safely. As an open and standard, ADL makes AI agents easier to build, review, and share in the open-source community.
• [Blog] AI Agent Engineering in Go with the Google ADK - AI, agents, and the related protocols touch on many open source projects. This post gives you a technical hands on with the Agent Starter Pack. By following it you'll learn how to build, test, and securely deploy a Go AI agent using Google Cloud services.
• [Article] How Kubernetes Broke the AWS Cloud Monopoly - Before Kubernetes, companies felt locked into AWS because of its unique APIs. Kubernetes allowed apps to run on any cloud, giving users more choice and helping other cloud providers grow. This has made multi-cloud the way forward for many enterprises. Are you utilizing a multi-cloud strategy? Has Kubernetes helped you get there?
• [Article] Even Linux Creator Linus Torvalds is Using AI to Code in 2026 - Opinions vary on where and whether AI is useful in various areas. One place that it has shown the greatest benefit is in as a tool for writing code. It seems Linus Torvalds has started to use it to assist with part of his AudioNoise side project. What a good way to find out how best AI can work for oneself. How have you been using AI with your code?

What exciting open source events and news are you hearing about? Let us know on our @GoogleOSS X account or our new @opensource.google Bluesky account.

A JSON schema package for Go

by Jonathan Amsterdam & Sam Thanawalla, The Go Team

JSON Schema is a specification for describing JSON values that has become a critical part of

LLM infrastructure. We recently released github.com/google/jsonschema-go/jsonschema, a comprehensive JSON Schema package for Go. We use it in the official Go SDK for MCP and expect it to become the canonical JSON Schema package for Google's Go SDKs that work with LLMs.

JSON Schema has been around for many years. Why are we doing this now, and what do LLMs have to do with it?

JSON is a flexible way to describe values. A JSON value can be null, a string, a number, a boolean, a list of values, or a mapping from strings to values. In programming language terms, JSON is dynamically typed. For example, a JSON array can contain a mix of strings, numbers, or any other JSON value. That flexibility can be quite powerful, but sometimes it's useful to constrain it. Think of JSON Schema as a type system for JSON, although its expressiveness goes well beyond typical type systems. You can write a JSON schema that requires all array elements to be strings, as you could in a typical programming language type system, but you can also constrain the length of the array or insist that its first three elements are strings of length at least five while the remaining elements are numbers.

The ability to describe the shape of JSON values like that has always been useful, but it is vital when trying to coax JSON values out of LLMs, whose output is notoriously hard to constrain. JSON Schema provides an expressive and precise way to tell an LLM how its JSON output should look. That's particularly useful for generating inputs to tools, which are usually ordinary functions with precise requirements on their input. It also turns out to be useful to describe a tool's output to the LLM. So frameworks like MCP use JSON Schema to specify both the inputs to and outputs from tools. JSON Schema has become the lingua franca for defining structured interactions with LLMs.

Requirements for a JSON Schema package

Before writing our own package, we took a careful look at the existing JSON Schema packages; we didn't want to reinvent the wheel. But we couldn't find one that had all the features that we felt were important:

1. Schema creation: A clear, easy-to-use Go API to build schemas in code.
2. Serialization: A way to convert a schema to and from its JSON representation.
3. Validation: A way to check whether a given JSON value conforms to a schema.
4. Inference: A way to generate a JSON Schema from an existing Go type.

We looked at the following packages:

• https://github.com/invopop/jsonschema provides inference, but not validation.
• https://github.com/santhosh-tekuri/jsonschema does not provide inference.
• https://github.com/xeipuuv/gojsonschema does not provide a way to construct a schema in code.
• https://github.com/qri-io/jsonschema does not provide inference or a way to construct a schema in code.

It didn't seem feasible to cobble together what we needed from multiple packages, so we decided to write our own.

A Tour of jsonschema-go

A Simple, open Schema struct

At the core of the package is a straightforward Go struct that directly represents the JSON Schema specification. This open design means you can create complex schemas by writing a struct literal:

var schema = &jsonschema.Schema{
  Type:        "object",
  Description: "A simple person schema",
  Properties: map[string]*jsonschema.Schema{
    "name": {Type: "string"},
    "age": {Type: "integer", Minimum: jsonschema.Ptr(0.0)},
  },
  Required: []string{"name"},
}

A Schema will marshal to a valid JSON value representing the schema, and any JSON value representing a schema can be unmarshalled into a Schema.

The Schema struct defines fields for all standard JSON Schema keywords that are defined in popular specification drafts. To handle additional keywords not present in the specification, Schema includes an Extra field of type map[string]any.

Validation and resolution

Before using a schema to validate JSON values, the schema itself must be validated, and its references to other schemas must be followed so that those schemas can themselves be checked. We call this process resolution. Calling Resolve on a Schema returns a jsonschema.Resolved, an opaque representation of a valid schema optimized for validation. Resolved.Validate accepts almost any value that can be obtained from calling json.Umarshal: null, basic types like strings and numbers, []any, and map[string]any. It returns an error describing all the ways in which the value fails to satisfy the schema.

rs, err := schema.Resolve(nil)
if err != nil {
  return err
}
err = rs.Validate(map[string]any{"name": "John Doe", "age": 20})
if err != nil {
  fmt.Printf("validation failed: %v\n", err)
}

Originally, Validate accepted a Go struct. We removed that feature because it is not possible to validate some schemas against a struct. For example, If a struct field has a non-pointer type, there is no way to determine whether the corresponding key was present in the original JSON, so there is no way to enforce the required keyword.

Inference from Go types

While it's always possible to create a schema by constructing a Schema value, it's often convenient to create one from a Go value, typically a struct. This operation, which we call inference, is provided by the functions For and ForType. Here is For in action:

type Person struct {
    Name string `json:"name" jsonschema:"person's full name"`
    Age int `json:"age,omitzero"`
}

schema, err := jsonschema.For[Person](nil)

/* schema is:
{
    "type": "object",
    "required": ["name"],
    "properties": {
        "age":  {"type": "integer"},
        "name": {
            "type": "string",
            "description": "person's full name"
        }
    },
    "additionalProperties": false
}
*/

For gets information from struct field tags. As this example shows, it uses the name in the json tag as the property name, and interprets omitzero or omitempty to mean that a field is optional. It also looks for a jsonschema tag to get property descriptions. (We considered adding support for other keywords to the jsonschema tag as some other packages do, but that quickly gets complicated. We left an escape hatch in case we decide to support other keywords in the future.)

ForType works the same way, but takes a reflect.Type. It's useful when the type is known only at runtime.

A foundation for the Go community

By providing a high-quality JSON Schema package, we aim to strengthen the entire Go ecosystem for AI applications (and, indeed, any application that needs to validate JSON). This library is already a critical dependency for Google's own AI SDKs, and we're committed to its long-term health. We welcome external contributions, whether they are bug reports, bug fixes, performance enhancements, or support for additional JSON Schema drafts. Before beginning work, please file an issue on our issue tracker.

Mentor Org Applications for Google Summer of Code 2026 open through Feb 3

by Stephanie Taylor, Mary Radomile & Lucila Ortíz, Google Open Source

Attention open source enthusiasts! Mentoring organization applications for Google Summer of Code (GSoC) 2026 are officially open. This is your opportunity to guide students and developers early in their careers. The application window begins today, Monday, January 19th, and will remain open until February 3, 2026, at 18:00 UTC.

To find more information about the process of becoming a mentor organization, please review our official GSoC site. We also recommend consulting the Mentor Guide and the GSoC Organization Admin Tips, as both provide tips for preparing your community and strengthening your application.

GSoC welcomes a wide variety of open source projects working in AI/ML, security, cloud, development tools, science, medicine, data, media, and more! For 2026, we are looking for even more innovative projects working on Artificial Intelligence/Machine Learning and Security.

Requirements for GSoC Mentoring Organizations:

1. An established open source project with at least 18 months of history
2. Software produced and released under an Open Source Initiative (OSI)-approved license
3. A robust community with members who are enthusiastic and prepared to mentor GSoC participants
4. An active project characterized by regular engagement, rather than infrequent contributions
5. A comprehensive list of Project Ideas (refer to the mentor guide for best practices)
6. A clear grasp of GSoC objectives and program rules
7. A high-quality application that provides a detailed explanation of your project and its specific goals

2026 Mentoring Organizations will be announced on February 19, 18:00 UTC*.

For first-time organizations interested in participating, we strongly suggest getting a referral from experienced organizations that think that your project is a good fit.

Google Summer of Code: Organizations Apply

Please visit the GSoC site for even more information on how to apply and review the detailed timeline for important deadlines this year. We recommend reading our help page on our website for easy access to all the most important resources for all the applicants.

We look forward to seeing your organization applications and learning more about your communities!

*Interested GSoC Contributor? After mentoring organizations are announced you can (and should!) begin researching each organization and reviewing project ideas to find the community that fits your interests. GSoC contributor applications are open from March 16-31.

Explore public datasets with Apache Iceberg & BigLake

by Talat Uyarer & Alex Stephen, Biglake Team

The promise of the Open Data Lakehouse is simple: your data should not be locked into a single engine. It should be accessible, interoperable, and built on open standards. Today, we are taking a major step forward in making that promise a reality for developers, data engineers, and researchers everywhere.

We are thrilled to announce the availability of high-quality Public Datasets served via the Apache Iceberg REST Catalog. Hosted on Google Cloud's BigLake, these datasets are available for read-only access to anyone with a Google Cloud account.

Whether you are using Apache Spark, Trino, Flink, or BigQuery, you can now connect to a live, production-grade Iceberg Catalog and start querying data immediately. No copying files, no managing storage bucket. Just configure your catalog and query.

How to Access Public Datasets

This initiative is designed to be engine-agnostic. We provide the storage and the catalog and you bring the compute. This allows you to benchmark different engines, test new Iceberg features, or simply explore interesting data without setting up infrastructure or finding data to ingest.

How to Connect with Apache Spark

You can connect to the public dataset using any standard Spark environment (local, Google Cloud Dataproc, or other vendors). You only need to point your Iceberg catalog configuration to our public REST endpoint.

Prerequisites:

• A Google Cloud Project (for authentication).
• Standard Google Application Default Credentials (ADC) set up in your environment.

Spark Configuration:

Use the following configuration flags when starting your Spark Shell or SQL session. This configures a catalog named bqms (BigQuery Metastore) pointing to our public REST endpoint.

PROJECT_ID=<YOUR_PROJECT_ID>

  spark-sql \
    --packages org.apache.iceberg:iceberg-spark-runtime-3.5_2.12:1.10.0,org.apache.iceberg:iceberg-gcp-bundle:1.10.0 \
    --conf spark.hadoop.hive.cli.print.header=true \
    --conf spark.sql.catalog.bqms=org.apache.iceberg.spark.SparkCatalog \
    --conf spark.sql.catalog.bqms.type=rest \
    --conf spark.sql.catalog.bqms.uri=https://biglake.googleapis.com/iceberg/v1/restcatalog \
    --conf spark.sql.catalog.bqms.warehouse=gs://biglake-public-nyc-taxi-iceberg \
    --conf spark.sql.catalog.bqms.header.x-goog-user-project=$PROJECT_ID \
    --conf spark.sql.catalog.bqms.rest.auth.type=google \
    --conf spark.sql.catalog.bqms.io-impl=org.apache.iceberg.gcp.gcs.GCSFileIO \
    --conf spark.sql.catalog.bqms.header.X-Iceberg-Access-Delegation=vended-credentials \
    --conf spark.sql.defaultCatalog=bqms

Note: Replace <YOUR_PROJECT_ID> with your actual Google Cloud Project ID. This is required for the REST Catalog to authenticate your quota usage, even for free public access.

Exploring the Data: Sample Queries

Once connected, you have full SQL access to the datasets. We are launching with the classic NYC Taxi dataset, modeled as an Iceberg table to showcase partitioning and metadata capabilities.

1. The "Hello World" of Analytics

This query aggregates millions of records to find the average fare and trip distance by passenger count. It demonstrates how Iceberg efficiently scans data files without needing to list directories.

SELECT 
    passenger_count,
    COUNT(1) AS num_trips,
    ROUND(AVG(total_amount), 2) AS avg_fare,
    ROUND(AVG(trip_distance), 2) AS avg_distance
FROM 
    bqms.public_data.nyc_taxicab
WHERE 
    data_file_year = 2021
    AND passenger_count > 0
GROUP BY 
    passenger_count
ORDER BY 
    num_trips DESC;

What this demonstrates:

• Partition Pruning: The query filters on data_file_year, allowing the engine to skip scanning data from other years entirely.
• Vectorized Reads: Engines like Spark can process the Parquet files efficiently in batches.

2. Time Travel: Auditing Data History

One of Iceberg's most powerful features is Time Travel. You can query the table as it existed at a specific point in the past.

-- Compare the row count of the current version vs. a specific snapshot
SELECT 
    'Current State' AS version, 
    COUNT(*) AS count 
FROM bqms.public_data.nyc_taxicab
UNION ALL
SELECT 
    'Past State' AS version, 
    COUNT(*) AS count 
FROM bqms.public_data.nyc_taxicab VERSION AS OF 2943559336503196801;

Description:

This query allows you to audit changes. By querying the history metadata table (e.g., SELECT * FROM bqms.public_data.nyc_taxicab.history), you can find snapshot IDs and "travel back" to see how the dataset grew over time.

Coming Soon: An Iceberg V3 Playground

We are not just hosting static data; we are building a playground for the future of Apache Iceberg. We plan to release new datasets specifically designed to help you test Iceberg V3 Spec features.

Start Building Today

The goal of these public datasets is to lower the barrier to entry. You don't need to manage infrastructure to learn Iceberg; you just need to connect. Whether you are a data analyst, data scientist, data engineer or a data enthusiast, today you can:

• Use BigQuery (via BigLake) to query these tables directly using SQL, combining them with your private data.
• Test your OSS engine (e.g. Spark, Trino, Flink etc.) configurations against a live REST Catalog.

Start building an open, managed and high-performance Iceberg lakehouse to enable advanced analytics and data science with https://cloud.google.com/biglake today!

Happy Querying!

This Week in Open Source #12

by Daryl Ducharme & amanda casari, Google Open Source

This Week in Open Source for January 9, 2026

A look around the world of open source

Here we are at the beginning of a new year. What will it bring to the open source world? What new projects will be started? What should we be focusing on? What is your open source resolution for 2026? One of ours is to better connect with various open source communities on social media. We've gotten off to a big start by launching an official Google Open Source account on Bluesky. Already, we are enjoying the community there.

Upcoming Events

• January 21 - 23: Everything Open 2026 is happening in Canberra, Australia. Everything Open is a conference focused on open technologies, including Linux, open source software, open hardware and open data, and the communities that surround them. The conference provides technical deep-dives as well as updates from industry leaders and experts on a wide array of topics from these areas.
• January 29: CHAOSScon Europe 2026 is co-located with FOSDEM in Brussels, Belgium. This conference revolves around discussing open source project health, CHAOSS updates, use cases, and hands-on workshops for developers, community managers, project managers, and anyone interested in measuring open source project health. It also shares insights from the CHAOSS context working groups including OSPOs, University Open Source, and Open Source in Science and Research.
• January 31 - February 1: FOSDEM 2026 is happening at the Université Libre de Bruxelles in Brussels, Belgium. It is a free event for software developers to meet, share ideas and collaborate. Every year, thousands of developers of free and open source software from all over the world gather at the event in Brussels.
• February 24 - 25: The Linux Foundation Member Summit is happening in Napa, California. It is the annual gathering for Linux Foundation members that fosters collaboration, innovation, and partnerships among the leading projects and organizations working to drive digital transformation with open source technologies.

Open Source Reads and Links

• [Talk] State of the Source at ATO 2025: State of the "Open" AI - At the end of last year Open Source Initiative gave a summary of Gabriel Toscano's talk at All Things Open. In the talk he discusses how AI models call themselves "open" but often lack the legal or technical freedoms that true open source requires. Analysis of ~20,000 Hugging Face models found Apache 2.0 and MIT are common, but many models have no license or use restrictive custom terms. The study warns that inconsistent labeling and mutable restrictions muddy openness and urges clearer licensing and platform checks.
• [Article] The Reality of Open Source: More Puppies, Less Beer - Bitnami's removal of popular containers last year shows that open source can suddenly change and disrupt users. Organizations must evaluate who funds and maintains each open source component, not just the code. Plan for business continuity, supply-chain visibility, and the ability to fork or replace critical components.
• [Blog] The Open Source Community and U.S. Public Policy - The Open Source Initiative is increasing its U.S. policy work to ensure open source developers are part of technology and AI rulemaking. Since policymakers often lack deep knowledge of open source, the community must explain how shared code differs from deployed systems. Joining groups like the Open Policy Alliance helps nonprofits engage and influence policy.
• [Article] Pebble, the e-ink smartwatch that refuses to die, just went fully open source - Pebble, the e-ink smartwatch with a tumultuous history, is making a move sure to please the DIY enthusiasts that make up the bulk of its fans: Its entire software stack is now fully open source, and key hardware design files are available too.
• [Article] Forget Predictions: Tech Leaders' Actual 2026 Resolutions - We want to know your open source resolutions and perhaps these resolutions from some tech leaders (open source and otherwise) can point you in a direction. Their plans run the gamut of securing and managing AI responsibly, reducing noise in security data, and creating healthier tech habits. The common theme is intentional, measurable change over speculation.
• [Paper] Everything is Context: Agentic File System Abstraction for Context Engineering - GenAI systems may produce inaccurate or misleading outputs due to limited contextual awareness and evolving data sources. Thus mechanisms are needed to govern how persistent knowledge transitions into bounded context in a traceable, verifiable, and human-aware manner, ensuring that human judgment and knowledge are embedded within the system's evolving context for reasoning and evaluation.
  
  The paper proposes using a file-system abstraction based on the open-source AIGNE framework to manage all types of context for generative AI agents. This unified infrastructure makes context persistent, traceable, and governed so agents can read, write, and version memory, tools, and human input.

What exciting open source events and news are you hearing about? Let us know on our @GoogleOSS X account or our new @opensource.google Bluesky account.

Training Marin 32B: What an open lab can build with TPUs, JAX, and a little persistence

by David Hall, Open Athena

Last summer, we partnered with Google to share how Marin trained a fully open 8B foundation model using JAX and TPUs. Since then, our process hasn't changed much, but the scale has. Over the summer, we trained a 32B model entirely in the open, and most days there was just one person keeping the run moving.

Large-scale training is usually associated with big teams and bigger infrastructure. Large model releases typically have hundreds of authors. Marin tests a different hypothesis: using open source software and data, small teams can train serious foundation models if the tooling is good, the platform is stable, and the process is transparent. The Marin 32B run was our strongest validation yet.

---

A model built with one hand on the helm

Marin was started at Stanford University's Center for Research on Foundation Models with the goal of building radically open foundation models. In May, we released Marin 8B Base, which bested the popular Llama 3.1 8B Base on 14 of 19 benchmarks. Marin 8B was trained using Google TPU v4 and TPU v5e from the TPU Research Cloud.

Building on that success, we set out to build a 32B model starting in June. Our 32B training run followed Marin's usual "Tootsie Roll" style: start with a solid recipe, instrument heavily, and adapt mid-flight when necessary. That flexibility matters, because the first time you train at a larger scale, issues inevitably arise.

The timing, however, was less than ideal, as universities tend to empty out over the summer. Students graduate, get internships, go home, or travel the world. Marin was no different. By June, our team was down to one full time research engineer, with a few PhD students providing guidance when they weren't busy with their dissertations. Nevertheless, we pushed forward.

To spoil the ending, the model turned out quite well. On release, Marin 32B Base was the best open source base model, and it outperformed comparable open-weights models like Google's Gemma 3 27B PT on 24 of 42 base-model evaluations.

There were many bumps along the way, resulting in multiple mid-run corrections, but through it all Google's TPU infrastructure stayed rock-solid, and JAX's predictable performance let us iterate quickly. This meant that even with a tiny team, we could diagnose, patch, and continue training without losing momentum.

To be blunt: one researcher kept the 32B run alive all summer, juggling preemptible slices, rebuilding optimizer state, switching architectures, and generally shepherding ~6.4 trillion tokens across v5p and v4 pods—while mostly working on other Marin projects. The fact that this was possible speaks to the stability of the TPU platform and the maturity of the JAX/Marin stack.

The short version of a long summer

Our retrospective goes into much more detail about every spike, switch and cooldown. Here's the condensed version.

We began with a Llama-3-style 32B backbone and our best 8B data mix, running on preemptible TPU v5p pods. Preemptions were predictable, and recovery was nearly automatic. As availability tightened, however, we moved to dedicated TPU v4 capacity. After a slight tweak to gradient checkpointing to accommodate the older hardware (made easy by JAX's built-in support), we were back up and running and performance stayed excellent.

Around 70k steps, persistent loss spikes appeared. We tried clipping, update-norm guards, skip-step heuristics, "necromancy" (rebuilding optimizer state), and swapping in optimizers like Muon. Nothing helped. The model needed architectural support.

So, we warm-started the run onto a Qwen3-style architecture, which is the same as the Llama 3 architecture, except that it adds QK-Norm to attention. After a brief loss bump, the spikes vanished. The model recovered to its expected trajectory within ~10 billion tokens and remained stable.

Towards the end of training, it was time for a cool down. When training LLMs, one "cools down" the model by lowering the learning rate and changing the data mix to higher quality data. Our first cooldown surfaced two issues: contamination from a cached math dataset, and a training-loss phase shift caused by our linear-congruential shuffle. Switching to a Feistel-based shuffle fixed the latter completely. After cleaning the data and re-running the cooldown, the second cooldown was smooth and produced the final model.

The result: a strong, open 32B base model

Marin 32B Base is a competitive open-source base model. It outperformed Olmo 2 32B Base—the previous best fully open-source base model—on 32 of 42 tasks, and it performs especially well on knowledge-heavy evaluations like ARC, BoolQ, and PIQA.

Head-to-head, Marin 32B Base also beat Gemma 3 27B PT on 24 of 42 tasks, and its overall average rank places it alongside Qwen 2.5 32B and the newer Olmo 3 32B models. On our evaluation suite, Marin 32B Base actually ties Olmo 3 32B Base in win rate, despite Olmo 3 being trained by a much larger team and arriving a month later.

Mean rank across our evaluation suite (lower is better). Marin 32B Base lands in the top cluster of open(-weight) models, alongside Qwen 2.5 and Olmo 3, and ahead of Gemma 3 27B PT and Olmo 2 32B. Gray bars indicate open weight models, while blue bars indicate open source models.

While Olmo 3 32B Base now comfortably leads on math and coding benchmarks, Marin 32B Base holds its own and still leads on many knowledge QA evaluations. For a model trained with a fraction of the team size typically expected for a 30B-scale run, we're proud of where it landed.

Because Marin 32B Base (like Olmo 3 32B) is open source, the weights, code, data recipes, and every experimental detour are public. Anyone can reproduce, audit, or build on the work.

---

The stack that made it possible

TPU stability across large slices

During the run, we moved across preemptible v5p-512 slices coordinated with Cloud TPU Multislice, a v4-2048 slice for the long middle, and several mid-run architectural transitions. Throughout, TPUs were completely reliable for us: no mysterious hangs, no collective-op debugging. Preemptions were predictable and easy to recover from.

JAX + Levanter = predictable performance

Levanter builds on JAX's XLA compilation. In practice, what mattered for us was deterministic restarts, stable MFU at scale without custom kernels, and JAX's activation checkpointing, which made the v5p to v4 migration easy.

Marin's experiment system

Marin logs every step of the experimental pipeline: hyperparameters, code versions, datasets, metrics, and artifacts. Even with architectural switches and restarts, the run never devolved into a tangle of scripts. And because it's all open, anyone can retrace or reproduce the training.

What's next

Marin 32B Base is a strong base model, but we're not done. Here's what's coming next:

• A reasoning-optimized Marin 32B
• Hardened multislice TPU support for smoother preemptible training
• Exploring MoE variants for the next scale
• Continuing to release everything, including successes and failures, openly

Closing thought

Training a 32B model with a small team isn't about heroics but about using the right tools and infrastructure. TPUs' reliability, JAX's clarity and performance, and Marin's open, reproducible process provided the leverage we needed. If the 8B run showed that open labs can build credible models, the 32B run showed they can do it at scale: quietly, steadily, and with far fewer people than you might expect.

SpatialReasoner: Teaching VLMs to "see" structure — Accelerated with Tunix on TPUs

by Yifan Shen & Ismini Lourentzou, University of Illinois Urbana-Champaign, and Srikanth Kilaru, Google ML Frameworks

Introduction

We are seeing an increasing interest in Tunix among researchers focusing on the post-training phase of model development. As a native JAX library, Tunix offers the flexibility needed to refine foundation models—including Vision-Language Models (VLMs) and not just LLMs—helping them significantly improve their spatial reasoning capabilities.

Today, we are highlighting the work of the PLAN Lab (Perception and LANguage Lab) at the University of Illinois Urbana-Champaign (UIUC). To address the critical lack of spatial awareness in VLMs, they built SpatialReasoner-R1, a model capable of fine-grained spatial logic. They utilized Tunix and leveraged the Google TPU Research Cloud (TRC) to scale their experiments.

In this blog, Professor Ismini Lourentzou and her team explain how they used Tunix's modular design to implement novel alignment algorithms and improve spatial reasoning in VLMs.

The "Where" Problem in VLMs

Modern Vision-Language Models (VLMs) can describe images and answer basic visual questions with impressive fluency. However, they often struggle with fine-grained spatial understanding. If you ask a VLM to estimate distances, directions, or the precise relative positions of objects, it frequently "hallucinates" coordinates or produces inconsistent reasoning with vague answers.

These capabilities are critical for real-world applications, such as robotics, where precise spatial reasoning enables safe and intelligent interaction with physical environments.

To bridge this gap, we developed the SpatialReasoner-R1 (4B and 8B versions), a model trained to perform step-by-step visually grounded spatial reasoning. It achieves 95.59 on Qualitative Accuracy and 77.3 on Quantitative Accuracy for our 8B fDPO model, outperforming the strongest baseline by ~9% in average accuracy on the SPATIALRGPT-Bench while preserving strong general vision-language abilities.

The Method: Fine-Grained Direct Preference Optimization (fDPO)

The secret sauce behind SpatialReasoner-R1 is a new technique called Fine-Grained Direct Preference Optimization (fDPO).

Standard alignment methods (like DPO) usually give a model a simple "thumbs up" or "thumbs down" for an entire response. But spatial reasoning is complex— for example, a model might correctly identify an object yet make a flawed logical inference about its location.

fDPO introduces segment-specific preference granularity. We optimize separate loss components for:

1. Descriptive Grounding: Does the model correctly perceive and describe the objects in the image?
2. Logical Reasoning: Is the step-by-step deduction sound and follows coherent spatial logic?

To generate high-quality training signals, we built a Multi-Model Monte Carlo Tree Search (M3CTS) data generation pipeline, which constructs diverse reasoning trajectories that guide the model toward reliable spatial understanding.

Tunix: Modularity for Novel Research

Implementing a custom objective like fDPO can be difficult in rigid frameworks. Tunix addresses this by providing a well-structured and extensible DPOTrainer that makes it possible to introduce new alignment objectives without reengineering the training pipeline.

This modularity meant we could reuse the entire underlying training stack—sharding, data loading, and loop management—while injecting our novel research logic with just a small amount of well-contained code.

While our backbone model (Sa2VA) required specific architectural handling, the core fDPO algorithm is model-agnostic. We found the Tunix experience smooth and well-documented, making it easy to prototype and iterate on fine-tuning workflows without reinventing the wheel.

Google TRC & TPUs: Reliability at Scale

Training a model to reason over long horizons requires significant compute. The Google TPU Research Cloud (TRC) provided the infrastructure we needed to make large-scale training practical.

• Scalability: Tunix's integration with TPUs allowed us to scale our experiments seamlessly.
• Reliability: The system performed reliably across multiple TPU runs, which was essential for conducting large-scale spatial reasoning benchmarks.
• Support: The Google Tunix and TRC teams assisted with infrastructure setup and experiment design, helping us refine our multi-model exploration strategy.

Looking Ahead: Open Source Contributions

We believe that open-source, extensible tools like Tunix are vital for fostering innovation. They lower the barrier for researchers to experiment with new training objectives without rebuilding core infrastructure.

In that spirit, we contributed our fDPO implementation back to the Tunix ecosystem. We open-source the core fDPO components, enabling the community to apply segment-specific preference optimization to their own models.

Get Started

You can explore our research and the tools we used below:

• SpatialReasoner Project Page
• Tunix Documentation
• Tunix GitHub Repository
• Google TRC
• JAX documentation
• JAX AI Stack documentation

GRL: Turning verifiable games into a post-training suite for LLM agents with Tunix on TPUs

by The GRL Team, UC San Diego and Lin Chai & Srikanth Kilaru, Google ML Frameworks

Introduction

JAX is widely recognized for its power in training large-scale AI models. However, a primary bottleneck in the next phase of AI development—LLM post-training with Reinforcement Learning (RL)—is the scarcity of environments with verifiable rewards.

Today, we are highlighting the work of the GRL (Game Reinforcement Learning) team at UC San Diego. To solve the data bottleneck, they have built a pipeline to turn video games into rigorous reasoning benchmarks. They utilized Tunix, a JAX-native research-friendly RL framework that supports multi-host, multi-turn capabilities, and leveraged the Google TPU Research Cloud (TRC) to scale their experiments. The results are promising: this approach has yielded significant improvements in model quality, particularly in planning and reasoning tasks, proving that games can be a viable substrate for serious AI capability training.

In this blog the GRL team explains how they are combining game environments, modular Tunix library for RL post-training, and TPU compute to train the next generation of agents.

---

Why Verifiable Games for LLM Post-Training?

Current RL post-training has shown strong gains in domains like math and coding because success can be auto-checked. However, these settings are often narrow and short-term. We are effectively overfitting RL to clean problems, while the next generation of agents must operate in messy, multi-step worlds.

To unlock RL as a systematic method for reasoning, we need a diverse pool of environments where rewards are grounded in explicit, machine-checkable rules. Games are this missing, underused substrate.

1. The Performance Gap: LLMs still perform surprisingly poorly on many strategy games, revealing a clear gap between model behavior and human-level interactive competence.
2. Verifiable Signals: Games come with built-in verifiable signals—wins, scores, puzzle completion—meaning outcomes are automatically and unambiguously graded without human labeling.
3. Long-Horizon Reasoning: Unlike short QA tasks, games force models to plan, explore, and reason over many steps.
4. Abundance: Decades of RL research has produced a standardized ecosystem of diverse environments ready to be recycled.

Game Reinforcement Learning (GRL): A Unified Game-to-Post-Training Pipeline

To harness this ecosystem, we built GRL, a comprehensive suite designed to recycle diverse game environments into a reusable post-training resource. Our mission is to prioritize environments with executable success checks—ranging from text-based puzzles to embodied 3D worlds and web/GUI workflows. Our code and ecosystem live under the LM Games organization (lmgame.org).

GRL provides three key capabilities:

• A Unified Pipeline: We standardize the conversion of games into RL-ready environments with structured states and consistent metrics. This makes results comparable across models and research groups.
• Versatile Configuration: Researchers can tailor interaction styles (e.g., max_turns, natural language feedback) while mixing training data from different tasks seamlessly. This allows for training on puzzles, math, and web tasks within a single run.
• Algorithm-Agnostic Interface: GRL works with any agentic training algorithm. While we frequently use PPO, the system serves as a robust testbed for developing new RL techniques.

The Engine: Plugging into the Tunix RL Framework

Designed for Research Flexibility and Multi-Turn Agents

In practice, plugging a GRL game agent into Tunix is seamless thanks to its modular design. Tunix is built specifically to support multi-turn agentic tasks, allowing researchers to leverage native one-turn inference APIs to achieve complex multi-turn rollouts, then batch those outputs directly back into the training flow. This research flexibility is key; the framework is lightweight enough for quick iteration and benchmarking, yet modular enough to allow fine-grained adjustments to reward functions, algorithms, and hardware-aware settings like mesh sizes.

We first define an agent_cfg (see picture above) that tells the system which game to play (eg. Sokoban or Tetris), how the LLM should talk (chat template + reasoning style), and its budgets (max turns, tokens per turn, action format). On the Tunix side, we then load a pre-trained model into three roles: actor, critic, and reference and build ClusterConfig to specify rollout and training configs and PpoConfig to specify RL hyperparameters. The glue is minimal and the layout is clear and research friendly: once agent_cfg, ppo_cfg, and cluster_cfg are defined, we construct an RLCluster and pass everything into PpoLearner, which gives us a complete multi-turn PPO trainer in JAX.

Our multi-turn RL workflow is equally lightweight from the user's point of view. For example, with a 5-turn budget, the trainer repeatedly lets the LLM "play" the game for up to five conversational turns: at each turn it sees the current grid or state, reasons in language using the chat template, outputs a series of actions, and receives the next state and a verifiable reward signal (win/loss/score/step penalty). GRL's agent + env configs handle all the orchestration: they log observations, actions, and rewards into structured trajectories, which Tunix then turns into token-level advantages and returns for PPO updates. You don't manually build datasets or rollouts; the trainer owns the loop - interact -> log -> compute rewards -> update policy -> repeat.

In our preliminary experiments using this setup, training Qwen2.5-7B-Instruct on Sokoban and Tetris yielded strong in-domain gains (+2-56% across game variants). We also observed modest generalization to out-of-domain tasks, with consistent improvements in planning tasks (Blocksworld: +3-7%) and positive but unstable signals in computer use (Webshop: ~+6%). All scripts and configs are available in the GRL repo: https://github.com/lmgame-org/GRL/tree/main. To reproduce the end-to-end Tunix + GRL training example (including our Sokoban/Tetris runs), you can simply clone the repo and run one line: bash tunix_quick_training_example.sh.

Google TRC & TPUs: Accelerating Game-Based RL at Scale

A critical component of our research was the Google TPU Research Cloud (TRC) program. Access to Cloud TPUs allowed us to move from small-scale prototypes to production-grade training runs with minimal friction.

TPUs and JAX directly attacked our two biggest bottlenecks:

1. Rollout Throughput: Using the vLLM-TPU path via tpu-inference, we could serve multiple model families on the same TPU v5p backend. This boosted sampling throughput, making the data-collection loop tighter and multi-environment concurrency cheaper.
2. Multi-Host Scale for 7B Models: Tunix's lightweight design combined with JAX's mesh-based sharding allowed us to scale the same code from a single host to multi-host setups declaratively. This capability was essential for our experiments with 7B parameter models (such as Qwen2.5-7B), where we leveraged 2 v5p-8 hosts with minimal code change (in fact, only an env var config). The scale up is seamless, proving that the infrastructure can handle the heavy computational lifting required for modern LLM post-training without requiring complex engineering overhauls.
3. Hardware Advantage: At the hardware level, the performance gains were significant. Each TPU v5p chip delivers around 459 BF16 TFLOPs, compared to roughly 312 on an NVIDIA A100. This raw power, combined with the TRC program's support, meant that large-N studies—involving more seeds, longer horizons, and more environments—became routine experiments rather than "special ops" engineering challenges.

This combination of Tunix's flexible abstraction and TRC's massive compute resources allowed us to iterate quickly on ideas while benefiting from production-grade infrastructure.

Get Started

GRL and Tunix are open for the community to explore. You can reproduce our end-to-end training example (including the Sokoban/Tetris runs) by cloning the repo, following the installation instructions, and then running a single command:

bash tunix_quick_training_example.sh

• GRL Repository
• Tunix Documentation
• Tunix GitHub Repository
• Google TRC
• JAX documentation
• JAX AI Stack documentation

ESCA: Grounding embodied agents with scene graphs — Accelerated by JAX

by The ESCA Team, University of Pennsylvania & Srikanth Kilaru, Google ML Frameworks

Introduction

Multi-Modal Language Models (MLLMs) are increasingly forming the core of the brain for general-purpose embodied agents — AI that can navigate and act in the physical world as robots. While MLLMs are making rapid progress, they often stumble on a critical hurdle: precise visual perception. They struggle to reliably capture the fine-grained links between low-level visual features and high-level textual semantics.

Today, we are highlighting the work of Prof. Mayur Naik's research team at the University of Pennsylvania. To bridge the gap between high-level language and low-level visual features, they developed ESCA (Embodied and Scene-Graph Contextualized Agent). By porting their neurosymbolic pipeline to JAX, they achieved the real-time performance necessary for high-throughput decision-making. This work also demonstrates that JAX drives performance gains across a wide range of hardware, including standard CPUs and NVIDIA GPUs, and not just on Google TPUs.

In this blog, the UPenn team explains how they combined structured scene graphs with JAX's functional design to reduce perception errors by over 50% and achieve a 25% speedup in inference.

---

The "Grounding" Problem in Embodied AI

Existing MLLMs are powerful, but they can be surprisingly "blind" when tasked with interacting with the physical world. In our empirical analysis of 60 navigation tasks from EmbodiedBench, we found that 69% of agent failures stemmed from perception errors. See the figure below.

The three top-level error types are Perception, Reasoning, and Planning. The second-level errors are Hallucination, Wrong Recognition, Spatial Understanding, Spatial Reasoning, Reflection Error, Inaccurate Action, and Collision. For clarity, the figure uses these acronyms to label the different error types.

The models struggle to capture fine-grained links between visual features and textual semantics. They might recognize a "kitchen," but fail to identify the specific spatial relationship between a knife and a cutting board required to complete a task.

Enter ESCA: The Anglerfish of AI

To solve this, we introduced ESCA, a framework designed to contextualize MLLMs through open-domain scene graph generation.

Think of ESCA like the bioluminescent lure of a deep-sea anglerfish. Just as the fish illuminates its dark surroundings to reveal prey, ESCA "illuminates" the agent's environment by generating a structured Scene Graph—a map of objects, attributes, and relationships (e.g., Cup [Red] ON Table).

A key innovation here is Selective Grounding. Injecting a massive scene graph of everything in the room can overwhelm the model. Instead, ESCA identifies only the subset of objects and relations pertinent to the current instruction. It performs probabilistic reasoning to construct prompts enriched with exactly the contextual details the agent needs to act.

The Engine: LASER and Scallop

At the core of ESCA is LASER, a CLIP-based foundation model trained on 87k video-caption pairs. LASER uses Scallop—our neurosymbolic programming language that supports JAX backends—to align predicted scene graphs with logical specifications. This pipeline allows us to train low-level perception models to produce detailed graphs without needing tedious frame-level annotations.

JAX User Experience

1. The Power of Statelessness

JAX's design encouraged a fully functional, stateless architecture. Every component, from feature extraction to similarity computation, was made into a pure modular function. This structure enabled effective use of jit (Just-In-Time) compilation. The XLA compiler could fuse sequences—like normalization, matrix multiplication, and softmax—into fewer kernels, reducing intermediate buffers and lowering GPU overhead.

2. Handling Complex Control Flow

Our pipeline requires selecting the "top-k" most relevant objects from a probabilistic scene graph. This introduces complex control flow. JAX provided the primitives we needed to handle this efficiently:

• We used jax.lax.cond to manage control flow inside the probabilistic graph.
• We leveraged jax.nn and jax.numpy for all activation functions and batched math in a JIT-friendly way.

3. Debugging and Transparency

Migrating to JAX was also a learning experience. Tools like jax.debug.print/callback() allowed us to inspect values inside jit-compiled functions, while jax.disable_jit() let us easily switch to eager execution to step through the program seeing intermediate values.

Furthermore, the transparency of the open-source system was impressive. Being able to read the annotated source code and see how Python functions trace into jaxpr (JAX expression) gave us deep insight into how to design inference logic that scales.

4. Seamless Integration with Flax

NNX fits into our workflow perfectly. We used nnx.Module to structure the model and FrozenDict to keep parameters organized and immutable. The TrainState object made managing model parameters and optimizer states straightforward, without adding the complexity often found in other frameworks.

JAX Performance: A 25% Speedup

Embodied agents operate in a continuous loop: planning, acting, and updating their understanding of a dynamic world. High latency here is a dealbreaker. We ported LASER from PyTorch to JAX to improve real-time performance, and the benefits were significant. By rewriting our core similarity computations and feature pipelines as pure functions wrapped in jax.jit, we achieved significant gains.

On an NVIDIA H100 GPU, JAX reduced the average time per frame from 18.15 ms (PyTorch) to 14.55 ms (JAX)—a roughly 25% speedup.

Framework	Hardware	Avg Time Per Frame (ms) ↓	FPS ↑
PyTorch	H100 GPU	18.15 ± 0.73	55.15 ± 2.31
JAX	H100 GPU	14.55 ± 0.64	68.82 ± 3.13

Conclusion

ESCA demonstrates that better data—structured, grounded scene graphs—can solve the perception bottleneck in Embodied AI. But it also demonstrates that better infrastructure is required to run these systems in the real world. JAX provided the speed, transparency, and modularity needed to turn our research into a real-time agent capable of reliable reasoning.

Acknowledgements

This research was made possible through support from a Google Research Award to the University of Pennsylvania and from the ARPA-H program on Safe and Explainable AI under award D24AC00253-00.

Get Started

You can explore the LASER code, the ESCA framework and documentation for JAX and Flax at:

• ESCA Repository
• LASER/SGClip Repository
• JAX Documentation
• Flax Documentation

Empowering app developers: Fine-tuning Gemma 3 for mobile with Tunix in Google Colab

by Henry Ndubuaku & Noah Cyclich, Cactus Compute and Lance Wang & Srikanth Kilaru, Google ML Frameworks

In the rapidly evolving world of AI models for mobile devices, a persistent challenge is how to bring SOTA LLMs to smartphones without compromising on privacy or requiring App developers to be Machine Learning engineers.

Today, we are excited to talk about how Cactus, a startup building a next-gen inference engine for mobile devices, fine-tunes the open-source Gemma 3 model. By leveraging Tunix, the LLM post-training library in the JAX ML ecosystem, they achieved this entirely on Google Colab's Free Tier.

The Challenge: Making Small Models "Expert"

For app developers, running Large Language Models (LLMs) in the cloud isn't always an option due to privacy concerns (like GDPR) and latency requirements. The solution lies in running models locally on the device. However, most smartphones globally lack specialized MPUs (Micro Processing Units), meaning developers need highly efficient, smaller models.

While compact models like Gemma (270M or 1B parameters) are incredibly efficient, they are often "generalists." To be useful for specific mobile applications—such as a medical imaging assistant or a legal document analyzer—they need to be fine-tuned to become domain experts.

The problem? Most app developers are not ML infrastructure experts. Setting up complex training pipelines, managing dependencies, and navigating steep learning curves creates too much friction.

The Solution: SFT via Tunix on Google Colab

To solve this, Cactus created a simplified "Low-Friction" workflow by implementing a Python script using Supervised Fine Tuning (SFT) APIs of Tunix in a Colab.

1. The Engine: Tunix

Cactus utilized Tunix, Google's lightweight and modular LLM post-training library, which supports both SFT and leading RL algorithms, and executes natively on TPUs. Tunix strips away the complexity of heavy frameworks, offering a simplified path to Supervised Fine-Tuning (SFT).

2. The Access: Google Colab Free Tier

Accessibility was a key requirement. Instead of requiring developers to set up complex cloud billing and project IDs immediately, the workflow operates entirely within a Google Colab Notebook. By utilizing the free tier of Colab, developers can:

• Load the Gemma 3 model.
• Upload their specific dataset (e.g., medical data or customer service logs).
• Run an SFT (Supervised Fine-Tuning) job using Tunix.
• Export the weights for conversion.

3. The Deployment: Cactus

Once tuned, the model is converted into the Cactus graph format. This allows the now-specialized Gemma 3 model to be deployed directly into a Flutter or native mobile app with just a few lines of code, running efficiently on a wide range of smartphone hardware.

Why This Matters

"Our users are app developers, not ML engineers," explains Henry Ndubuaku, co-founder of Cactus. "They want to pick a model, upload data, and click 'tune.' By using Tunix and Colab, we can give them a 'clone-and-run' experience that removes the intimidation factor from fine-tuning."

This workflow represents the "lowest hanging fruit" in democratizing AI:

• No complex local environment setup.
• No upfront infrastructure costs.
• High-performance JAX native Tunix library to tune a leading OSS model (Gemma).

What's Next?

While the Colab notebook provides an immediate, accessible solution, Cactus is exploring a future plan to build a full GUI-based portal for fine-tuning and quantization of LLMs with the back end compute as Google Cloud TPUs, allowing for scalable training of larger models and even more seamless integration into the mobile development lifecycle.

Get Started

Ready to turn your mobile app into an AI powerhouse? Check out the Tunix SFT Notebook for Cactus and start fine-tuning Gemma 3 for your device today:

• GRPO Notebook
• QLoRA Notebook
• DPO

You can explore Tunix sample scripts, documentation and repo at:

• Tunix sample scripts
• Tunix Documentation
• Tunix GitHub Repository

Shape the future with Google Summer of Code 2026!

By Stephanie Taylor, Mary Radomile & Lucila Ortiz

Are you a passionate beginner ready to make your mark in open source? Now is your chance to make an impact in the 2026 Google Summer of Code (GSoC) program!

For over two decades, GSoC has been a critical launchpad, consistently introducing fresh, enthusiastic talent into the open source ecosystem. The numbers speak for themselves: GSoC has connected more than 22,000 contributors with over 20,000 mentors from 1,000+ open source organizations, all collaborating to keep this vital community thriving.

---

Google, like everyone else, is built on open source. We depend on a healthy, vibrant open source ecosystem and want to lower the barrier to entry for people who want to contribute to the open community. Join in, learn great skills, and make an impact on people around the world.

Richard Seroter, Chief Evangelist, Google Cloud

---

Over 3+ months, contributors spend their time collaborating on real-world projects right alongside experienced mentors and their communities. This isn't just about coding; this deep immersion in open source does more than build valuable technical skills. It cultivates a strong understanding of community dynamics, best practices, and the soft skills needed to become a truly impactful open source contributor.

Check out the trailer below to learn more about the massive impact GSoC has made over the last two-plus decades.

Google Summer of Code trailer

Join GSoC as a Mentoring Organization

Application Period: January 19 – February 3

If you're interested in your organization participating, now is the time to start! We welcome around 30 new organizations to GSoC annually—and yours could be one of them.

• Visit our website where you'll find supportive materials to get started.
• The Mentor Guide is a must-read, offering an introduction to GSoC, community engagement tips, project idea suggestions, and guidance on applying.
• Pro Tip: For 2026, we'll have an expanded focus on the AI, Security, and Machine Learning domains.

Want to be a GSoC Contributor?

Application Period: March 16 – March 31

Ready to get started? The official website is your first stop!

• Watch the Introduction to GSoC video for a quick overview.
• For more in-depth preparation, read the Contributor Guide—it's packed with insightful ideas from past participants.
• Be sure to review the program rules and FAQ to understand eligibility requirements and how the program works.

---

Spread the Word!

Please help us amplify the message about GSoC 2026! Share this post with your peers, family members, colleagues, universities, and anyone interested in making a real difference in the open source community.

Join us and help shape the future of open source!

AI Conformant Clusters in GKE

AI Conformant Clusters in GKE

by Janet Kuo & Federico Bongiovanni, GKE

We are excited to announce that Google Kubernetes Engine (GKE) is now a CNCF-certified Kubernetes AI conformant platform, designed to provide a stable and optimized environment for your AI/ML applications. This initiative, culminating in a major announcement of the Kubernetes AI Conformance program by CNCF's CTO Chris Aniszczyk, at KubeCon NA 2025, is set to simplify AI/ML on Kubernetes for everyone. You can check out the Opening Keynote here.

During the keynote, Janet Kuo, author of this blog and Staff Software Engineer at Google, performed a live Demo, demonstrating the practical power of an AI Conformant cluster. If you are interested in the technical specifics, you can learn more about the demo here.

Why AI Conformance Matters

The primary goal of the Kubernetes AI Conformance program is to simplify AI/ML on Kubernetes, guarantee interoperability and portability for AI workloads, and enable a growing ecosystem of AI tools on a standard foundation.

Setting up a Kubernetes cluster for AI/ML can be a complex undertaking. An AI-conformant platform like GKE handles these underlying complexities for you, ensuring that your environment is optimized for scalability, performance, portability, and interoperability.

For a detailed look at all the requirements and step-by-step instructions on how to create an AI-conformant GKE cluster, we encourage you to read the GKE AI Conformance user guide.

What Makes GKE an AI-Conformant Platform?

A Kubernetes AI-conformant platform like GKE handles the underlying complexities for you, providing a verified set of capabilities to run AI/ML workloads reliably and efficiently. Here are some of the key requirements that GKE manages for you:

• Dynamic Resource Allocation (DRA): GKE enables more flexible and fine-grained resource requests for accelerators, going beyond simple counts. This is crucial for workloads that need specific hardware configurations.
• Intelligent Autoscaling for Accelerators: GKE implements autoscaling at both the cluster and pod level to ensure your AI workloads are both cost-effective and performant.
  • Cluster Autoscaling works at the infrastructure level. It automatically resizes node pools with accelerators, adding nodes when it detects pending Pods that require them and removing nodes to save costs when they are underutilized.
  • Horizontal Pod Autoscaling (HPA) works at the workload level. HPA can automatically scale the number of your pods up or down based on real-time demand. For AI workloads, this is especially powerful, as you can configure it to make scaling decisions based on custom metrics like GPU/TPU utilization.
• Rich Accelerator Performance Metrics: GKE exposes detailed, fine-grained performance metrics for accelerators. This allows for deep insights into workload performance and is essential for effective monitoring and autoscaling.
• Robust AI Operator Support: GKE ensures that complex AI operators, such as Kubeflow or Ray, can be installed and function reliably, enabling you to build and manage sophisticated ML platforms with CRDs.
• All-or-Nothing Scheduling for Distributed Workloads: GKE supports gang scheduling solutions like Kueue, which ensure that distributed AI jobs only start when all of their required resources are available, preventing deadlocks and resource wastage.

A Unified and Evolving Standard

The Kubernetes AI Conformance program is designed as a single, unified standard for a platform to support all AI/ML workloads. This reflects the reality that modern AI processes, from training to inference, increasingly rely on the same underlying high-performance infrastructure.

What's Next?

We invite you to explore the benefits of running your AI/ML workloads on an AI-conformant GKE cluster.

• Learn more about the program: check out the CNCF Kubernetes AI Conformance repository.
• Dive deeper into AI/ML on GKE: read the Introduction to AI/ML workloads on GKE.

The launch of the AI Conformance program is a significant milestone, but it is only the first step. We are eager to continue this conversation and work alongside the community to evolve and improve this industry standard as we head into 2026.