We are excited to announce that SkyRL now implements the Tinker API. Run Tinker training scripts on your own hardware with zero code changes. Try it out today: novasky-ai.notion.site/skyrl…

How does one RL post-train a 397B model for long-horizon knowledge work? 👩‍💼 We share every step we took to bring Qwen 3.5 397B from 16.1% Pass@1 to 27.3% on APEX-Agents using DPPO, including final models weights and the full training script🚀 This is the first of many works from Mercor Research on open model training research. Full blog: mercor.com/blog/training-fro… Source code: github.com/Mercor-Intelligen…

Worked with Mercor Research and the SkyRL team, training Qwen3.5-397B-A17B on APEX-Agents (long-horizon office work) off-the-shelf data with SkyRL, improving Pass@1 from 16% to 27%. The post is more of a practical field guide for what to do given an RL dataset, de-risking step by step: environment infra, harness fixes/optimizations, token-in-token-out, an overfitting run, small-model ablations, then the hero run. We were also delighted to see the gains transfer to OpenCode despite training on an MCP-based harness, and even to Terminal-Bench 2.1. Code, models, and eval traces are all open source. Hope it helps your next hillclimb!

In RL training, a vLLM rollout engine and a Megatron trainer can run the same policy yet disagree on a token's logprob due to floating-point non-associativity. SkyRL's IsoExec combines an execution contract with a unified model, aligning rounding-sensitive execution choices across rollout and training. Bitwise parity holds across different TP, EP, and SP layouts. For Gated DeltaNet, the chunkwise-parallel recurrent algorithm makes parallel training and prefill bitwise identical to recurrent decode. Qwen3.5-35B-A3B, DAPO, 8xH100, 50 steps: logprob diff 1.6e-2 to 6.7e-7, full-step overhead 25.3% ✅ vLLM's scheduler and CUDA graphs still apply. Thanks to @JiangAlexander1 and the SkyRL team at @NovaSkyAI. 🔗 vllm.ai/blog/2026-08-21-isoe…

It was a packed house at the SkyRL Meetup today. Thanks to all the speakers!

The first-ever SkyRL Meetup is happening! A year of hard work from the team, and it's been great to see users pick it up from across industry and academia. Join us Aug 18 Tuesday 6:00pm in SF for four lightning talks from enterprise users and collaborators (@trajectorylabs , @datadoghq , @AMD , @mercor ), followed by a SkyRL technical deep dive. Would love to see you there!

Join us for the first-ever SkyRL Meetup! It's been a year since we released the SkyRL library. SkyRL now has 2k+ GitHub stars and a growing user base in academia and industry, including researchers from Stanford, CMU, UC Berkeley, Microsoft AI, Datadog, and more. We're hosting a SkyRL Meetup on August 18th at the Anyscale office in SF. We have a special set of lightning talks: 🔗 Making frontier RL training accessible to everyone, from @trajectorylabs 🐶 Post-training an SRE agent with SkyRL, from @datadoghq ⚡️ Scaling RL with SkyRL on AMD GPUs, from @AMD 🧑🏻‍💻 Training knowledge work agents with SkyRL, from the SkyRL team (@charlie_ruan) We’ll also do a deep dive into the library and talk about the road ahead. More details here: luma.com/y4ahouh6

Excited to announce the release of SkyRL v0.3.0! SkyRL’s focus in this release revolves around three major themes: 1. Support for large scale, async RL 2. Improving SkyRL’s Tinker API for self-hosting Tinker servers 3. Providing a unified library for SFT and RL Some highlights 🧵

Today, we're excited to share that Biomni is published in @ScienceMagazine. Biomedical research is still fragmented, manual, and difficult to scale. In this work, we introduce Biomni - the first general-purpose biomedical AI agent with an integrated biology environment that can reason, plan, and execute end-to-end scientific workflows. We show that, with the right environment and harness, AI can automate large-scale omics analyses, orchestrate laboratory robotics, optimize molecular properties, and even train new AI models for biology. We also introduce a reinforcement learning recipe for continually improving biomedical AI agents, enabling open-source models to achieve frontier-level performance. It's surreal to look back. We started the Biomni project in early 2024, when agentic AI was still nascent. It is exciting to see tens of thousands of biologists collaborating with agents every day to accelerate science. Try Biomni: biomni.phylo.bio Read more: science.org/doi/10.1126/scie… This work is not possible without this truly inter-disciplinary team: @serena2z @hcwww_ @YuanhaoQ Minta Lu, Ryan Li, @yusufroohani Lin Qiu @shiyi_c98 Gavin Junze Di @rickwierenga @kavi_deniz Sherry @TianweiShe Shruti Jennefer Xin Zhou @MWheelerMD Jon Bernstein @MengdiWang10 @PengHeAtlas @zhou_jingtian @SnyderShot @lecong Aviv Regev @jure @StanfordAILab @genentech @phylo_bio @arcinstitute @UW @berkeley_ai @RetroBio_ @tamarindbio @Princeton @UCSF

Biomni is now out in Science @ScienceMagazine ! Huge congrats to the team 🎉 Biomni is a really exciting step toward AI agents that can carry out real biomedical research workflows. Proud that SkyRL supported the RL training for Biomni-R0 early last year, where we explored end-to-end multi-turn RL for biomedical reasoning agents. It’s really exciting to see how AI agents have evolved the past one year toward carrying out real scientific workflows and accelerating discovery.

SkyRL Tinker implementation is so cool, but still early days so I went ahead and implemented a basic chunked cross-entropy for 23x less memory usage! Now most models fit much more seamlessly in memory when combined with Megatron.

In parallel we announce a new open source repo: github.com/Snowflake-AI-Rese… This is the framework for very fast RL (and future other optimizations rolled into it) It currently has all the code you need to use or integrate Arctic RL into RL frameworks, with SkyRL and Verl available and more framework integrations coming. Please kindly spread the word! Thank you!

After many months of intense work the @Snowflake AI Research team is happy to present to you the new open source project: Arctic RL snowflake.com/en/blog/engine… - Arctic RL integrates with VeRL and SkyRL today; enable ZoRRo with one config flag, no code changes required - ZoRRo delivers up to 6x actor-update acceleration and a 3.5x end-to-end training speedup, reducing Arctic-Text2SQL-R2 training from ~5 days to ~36 hours on 32 H200 GPUs - Arctic-Text2SQL-R2 achieved higher accuracy scores (48.7) than Gemini 3.1 Pro (47.9) and Claude 4.7 (47.3) on Snowflake's evaluated enterprise SQL benchmark under the tested conditions - Two open source recipes ship with this release: a text-to-SQL recipe that improved BIRD dev accuracy from 59.92% to 70.35%, and a multi-hop QA recipe that improved average accuracy from 69.6% to 72.3%

🏹5 Days of Trajectory. Day 3 - An Open Source Training Stack for Continual Learning Building the platform for continual learning requires both partnering with pioneering AI companies, as we showed on Day 2 with Harvey, and working toward frontier research, which we are highlighting today. Continual learning means models that improve hourly from real production use. But with the size of frontier models, this becomes quite difficult. A Qwen-397b would need to spin up and tear down repeatedly across six GPU nodes, and that's valuable time gone. Our contribution is Continual LoRA (C-LoRA): many lightweight adapters running at once on one shared base model. Our insight centers on where the parallelism lives: instead of splitting one giant job across nodes, we load-balance many small jobs over a single base. The result: 2.81x experiment throughput over single-tenant training, with no regression on rewards. We built this together, with @anyscalecompute, @NovaSkyAI, and generous support from @GoogleCloud and @GoogleStartups. We've open-sourced on SkyRL as one of the first multi-LoRA, RL training platforms, so that every team can get to continual learning faster. We’re very excited to see what you build, please reach out!

We've shipped two major upgrades for RL✨! 1. Native weight syncing APIs: Standardizes weight transfer, provides optimized implementations for NCCL and CUDA IPC out of the box, and also lets frameworks easily bring their own. 2. Improved pause/resume for Async RL: Careful coordination between DP ranks so that engines don’t deadlock. Validated at scale in P/D, wide-EP setups! In collaboration with @anyscalecompute, @NovaSkyAI, and @RedHat. More and more RL frameworks are using vLLM as the default for inference, details in the blog 👇 vllm.ai/blog/2026-05-28-nati…

Amazing work! More and more RL frameworks are using vLLM as default. @vllm_project along with @anyscalecompute and @NovaSkyAI revamped weight syncing and improved wide-ep deployment for rollout!

it's been a really great experience working w/ @j316chuck and the @trajectorylabs team on building out their post-training stack for continual learning on top of SkyRL really excited to continue collaborating and seeing how the team can push the frontier for continual learning!

Awesome to see Trajectory labs launch out of stealth! It's been great collaborating with them in building out multi-LoRA for SkyRL!

RT @charlie_ruan: Excited to have supported @trajectorylabs with the SkyRL team over the past month, bringing training onto their own clust…

Today, @MichaelElabd, @QuantumArjun, and I are excited to announce Trajectory. We are a research lab and product company building the platform for Continual Learning. Our platform unlocks the signal already sitting in product usage, so companies can continuously post-train large-scale agentic models that outperform the frontier. @trajectorylabs We’ve raised $15M from @Conviction, @BessemerVP, @radicalvcfund, @jeffdean, @drfeifei and more. We’re partnering with some of the best AI-native companies: @ClayRunHQ @Harvey, @DecagonAI, @mercor, @RogoAI to power their agentic systems, some of which we are already in production with. We’ve brought together a world class research team from DeepMind, OpenAI, Apple, Meta Superintelligence, Amazon AGI, Scale AI, and an elite product team from Stripe and Figma. AI will never again start on day one. Every correction, every retry, every edit will make products smarter. This is Continual Learning.

Super cool work built with SkyRL! Have also thought about how it's so wasteful not to train on the observation tokens. Can't wait to upstream it:)