Qwen3.8-27B at 144 tok/s on an M5 Max MacBook Pro ⚡ Meet Inco Splash: our open-source inference engine, built around the model and around Apple silicon. Up to 3× the decode speed of Ollama, 2× oMLX, and almost 4× when an agent fans out into sub-agents.

We're adding support for AGENTS.md to Claude Code. Starting today in version 2.1.277, if there is no CLAUDE.md in a folder, Claude will check for and use AGENTS.md. You can toggle this behavior in /config.

Today, we’re announcing Ternary Bonsai 2 27B. Based on Qwen3.8 27B, Bonsai 2 27B is 9x smaller than its full-precision counterpart while retaining 98.2% of its aggregate benchmark performance. Two months after the first Bonsai 27B release, the biggest change is quality. The footprint remains 5.9 GB, but the gap to full precision has narrowed materially, with particularly strong gains in agentic coding, multimodal reasoning, and long-horizon tool use. Ternary Bonsai 2 27B is available today under Apache 2.0.

@georgebsocial If one made this as a movie, I think today's world models would be perfectly capable of making it "locally playable" like this without allowing much deviation from the movie. This would be doable at AAA levels long before a long-duration open world game. @DLeitersdorf

As a kid, I would've killed to be Ash Ketchum. So I built a game where you get to be Ash in VR. Do anything you can imagine, from petting Growlithe to blasting Team Rocket. The world sees it and generates around me in real time using @fal H3 Max

I’m speedrunning a company with @poteto and @roshan_s next week we’ll go from nothing to a business in 72 hours and stream the whole thing we don’t have a name, a product, or an idea yet we’ll use @bot as our employees, but keep real people (us) at the core Lauren says we’ll be turned into pizza in some sort of bond-villain punishment if we fail watch:

Evan and I were debate partners in high school; in a weird twist of fate, he’s now an alignment lead at Anthropic and I’m a researcher at OpenAI. We should not preserve the status quo. Humanity faces a series of challenges, independent of AI, that compel us to innovate. AI does pose risks to society. These threats are more ill-defined, and more probabilistic, than knowing that 16% of our loved ones will die from cancer, or that 50% of the currently inhabited regions on Earth will become too hot to sustain human life by 2070. While we ought soberly evaluate the risks of super-intelligence – and work extremely hard to mitigate those risks – we hold an equal moral obligation to paint the inspirational and optimistic vision of the world we want to create. As someone who worked on climate tech for 3 years prior to joining OpenAI, I have a different perspective on the role that AI must play in our future. I believe in empowering humanity with super intelligence that can build the world we want to live in: rebuilding polar ice caps, inventing new desalination techniques, electrifying the grid; printing solutions to pathogens in real-time; democratizing education worldwide.