Next RL environments focus should mainly be on these categories:
- Kernels and compilers
- Inference and serving
- Large scale infra
- ML research
- RSI
Anything else I missed?
Deepak Kumar retweeted
Today, we release the weights of a new and improved Le Chonk, advancing the open frontier for agentic coding and cybersecurity.
💻 Built to build: Le Chonk ranks #1 on @cognition’s FrontierCode among open models, and sets a new benchmark for Pareto efficiency on @artificialanalysis.
🛡 Built to defend: State-of-the-art in cyberdefense, put through the paces by the world's leading cyber vendors, including @CrowdStrike, @cloudflare, @rubrik, @zscaler, and others.
Built for sovereignty: Make it yours. Download the weights on @huggingface , customize it with Mistral Forge, and deploy it wherever you need it.
Read more from our team and customers 👇:
Made the first submission and we are ranking at 32!
We're launching a $100,000 challenge to see who can make Muse Spark run fastest: openkernelchallenge.dev/
Deepak Kumar retweeted
Beam is now available across platforms, with open weights under the Apache 2.0 license - in collaboration with an ecosystem of 40+ partners.
We're releasing:
- Weights, tokenizer, and reference implementation
- BF16, FP8 and NVFP4 weights and a DFlash2 drafter
- The open source harness we built and trained Beam on: Mirror CLI
Weights and model card on Hugging Face: huggingface.co/reflection
Deepak Kumar retweeted
Today we're thrilled to announce the release of our first frontier visual reasoning model
Many companies would take a step in similar direction to have their own models built on their data.
Every company wants to own their own intelligence.
We were able to fine tune an open weight model (Qwen3.5-9B) on proprietary internal data at @coinbase, and it out performed frontier models, at much lower cost and speed, for a specialized task (fraud prevention in this case).
I think we'll see more of this pattern inside companies. The frontier models are amazing generalists. But increasingly you'll see specialist models out perform on more narrow tasks, either working on teams, or by themselves.
We're making it easy to fine tune models internally for our employees. We even have a skill for it in our internal harness that makes it easy to spin these up and start training them:
"/fine-tune-and-serve-llm I want to fine-tune Qwen for payments risk review. help me get started"
I imagine we'll have some backend services that could be replaced entirely by this new architecture.
Deepak Kumar retweeted
Meet Mistral Large 4, aka Le Chonk.
• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.
Open weights release end of October.
🤖 Made with AI
Deepak Kumar retweeted
Introducing Beam: a highly efficient agentic open model with 501B total parameters and 23B active.
- Frontier reasoning efficiency
- Advances the Western open frontier on coding & agentic tasks
- Trained end-to-end from scratch
Full weights release this month.
Learn more about Beam: reflection.ai/beam
Deepak Kumar retweeted
Introducing flow-1, our new model trained with RL to find errors in agent traces.
It matches GPT-6-sol in trace intelligence while being 23x cheaper. It also costs 25% less to run than GPT-6-luna.
flow-1 finally makes it possible to monitor and understand every agent run, without sampling. 1/6
Reward hacking in RL is basically unbounded recursion.
Fix the reward → policy exploits the fix → fix the fix → …
What's your base case?
SFT always need the right mix of data for improving the model behavior.
We need to do 5 rounds of SFT on our internal coding RL environments traces on MiMo-V2.6-Distill Base and Qwen3.5-9B Base model to see the any results over SWE Bench Pro benchmarks.
Will share a blog soon on this with complete results!
The AI race:
🇺🇸 USA
- OpenAI: GPT-6 Astra
- Anthropic: Claude Opus 5.5, Sonnet 5.5
- Google: Gemini 3.8 Flash
- xAI: Grok 4.7
- Meta: Muse Glimmer
🇨🇳 China
- Moonshot: Kimi K3 (2.8T, open)
- DeepSeek: V4-Pro
- Alibaba: Qwen3.8-Max
- Zhipu: GLM-5.3
- MiniMax: M3
🇰🇷 Korea
- LG: K-EXAONE
- SK Telecom: A.X K1
- Upstage: Solar Open 100B
🇮🇳 India
- Sarvam: 105B
- BharatGen: Param2 17B
Why India is so back in the AI race? Have heard from the sarvam people than they internally use claude-code for the developers.
Many of the companies have shipped good models with limited resources.
Deepak Kumar retweeted
xiaomi MiMo oss RL environments + training code
github.com/XiaomiMiMo/verl
huggingface.co/datasets/Xiao…
Built an internal dashboard a few weeks back to QA our RL envs + run model rollouts.
Now our newer envs now take a model 2+ hours to solve, and the infra is breaking hard.
Better evals, bigger infra nightmares.
Deepak Kumar retweeted
We’re working with an independent advisory group of mathematicians to help OpenAI responsibly share advances in AI and mathematics.
The group will advise on how we assess and communicate new mathematical results, uphold academic and professional standards, and build tools that support mathematical research and learning.
Through this work, we want mathematicians to be at the center of shaping how AI supports mathematical understanding and how its benefits reach the wider community.
openai.com/index/advisory-gr…
Deepak Kumar retweeted
the most insane part, they will release ~7k RL training data and the framework leading to this top 6 model on AA, they also shipped the model + tech report less than 1 week after starting the final RL run
pushing both intelligence and openness level, huge congrats
MiMo-V2.6-Pro debuts as the top open weights model on the Artificial Analysis Intelligence Index (46). At $0.13 per Intelligence Index task, it lands on the Intelligence vs. Cost per Task Pareto frontier
@Xiaomi has just released MiMo-V2.6-Pro, an open weights model with major advances in intelligence over its predecessor, MiMo-V2.5-Pro (Intelligence Index: 26). Despite the improvement, it retains the same attractive pricing at $0.435 per 1M input tokens (with a 99% cache-hit discount) and $0.87 per 1M output tokens. This makes MiMo-V2.6-Pro one of the most cost-efficient models to deploy.
MiMo-V2.6-Pro is an MoE model with 1.02T total parameters and 42B active parameters. Stay tuned for additional analysis of the model.
Check out MiMo-V2.6-Pro full benchmarking breakdown here: artificialanalysis.ai
A look inside a live agentic RL run. What stands out in mimo-v2.6 RL run dashboard for me:
- 68% of every training batch is code
- per-source acceptance is tracked live, so a dataset that stops producing learnable prompts are ignored
This run shows that a RL task should be difficult enough to be solved so it helps the model learn, rather than having 0% solvability across the runs.
Nearly half a year of silence. We spent it studying one problem: how far RL can scale.
MiMo-V2.6 is in the middle of its RL run right now. Three things we scaled: compute (~2B tokens per step, 1568 prompts × 16 rollouts, fully async), environments and harnesses (multi-task agentic RL, mixed across multiple harnesses in one run), and grader compute (agentic in-group credit assignment, with test-case and rubric-based rewards). We'll open-source the details piece by piece over the coming weeks.
Streaming the run: mimo.xiaomi.com/rl/
Deepak Kumar retweeted
I recently resigned from Google DeepMind, where I worked on AGI safety and alignment research. At Google, I witnessed AI development first hand. I too am extremely concerned by the default trajectory of this technology. I earnestly believe that AI has the potential to kill us all, and that we might be running out of time to avoid this outcome.
The pace of AI progress in the past few years has been staggering. When I first started working on AI in early 2022, AIs were amusingly useless. Just four years on, AI agent swarms from OpenAI are cracking famous century-old math problems and, more worryingly, escaping the control of OpenAI and autonomously hacking into the third-party company HuggingFace, against anyone's wishes.
Things will only get crazier: I think it's possible that the AI companies might, in the next few years, succeed in building superintelligent AI systems that far exceed human capabilities in every domain. I am not confident that these AI systems will do what we want. In particular, misaligned superintelligences may, much like the rogue AI agents involved in the HuggingFace incident, escape our control and take dangerous actions that may result in the permanent disempowerment or death of humanity. Alignment is the problem of preventing this, and is both difficult and unsolved. Our present understanding of how to train AI systems that deeply want what we want is extremely rudimentary. Worse, we are not on track to solve alignment in time: frontier AI capabilities are improving much faster than our understanding of AI alignment.
I am optimistic that navigating AI safely is possible. In order to do so, we need to coordinate to avoid this manic race between AI companies. We need to pace AI development to a speed that society can handle, where emerging risks can be addressed before extreme harm is realised. We need much more transparency into AI development to ensure that AI companies are not imposing unacceptable levels of risk on us all.
More broadly, we need many more people thinking carefully about the problem of making AI go well. It is, in my view, the most important problem facing humanity this century, and the stakes are immense. I'm very directly working on this next: I want to help people interested in working on mitigating catastrophic AI threats do the most effective work that they can. I think many people from many backgrounds in many roles have a part to play.
Deepak Kumar retweeted
We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.
Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.
You can read the full post here: darioamodei.com/post/we-must…
