General
- Anthropic confidentially submits draft S-1 to the SEC. This suggests they are considering an IPO.
- Anthropic CEO Dario Amodei publishes: Policy on the AI Exponential.
Safety
- In response to the US White House Executive Order on AI, OpenAI release a blueprint: Democratic Governance of Frontier AI: A blueprint for a federal framework.
LLM
- Anthropic expand access to Mythos to more select companies.
- Anthropic releases Claude Fable 5, a “Mythos-class” model. Extremely capable.
- Google release Gemma 4 12B, an open model with multimodality (text, image, video, audio).
- Google release DiffusionGemma as open-source. Using text diffusion to generate blocks all-at-once (and then refine), they show 4× speed boost.
- Anthropic introduces Claude Tag, a slack-bot that makes Claude operate more like a coworker.
- OpenAI announce GPT-5.6 Sol.
- Anthropic announce Claude Sonnet 5.
Multi-modal
Agents
- Qwen announces Qwen-AgentWorld, a native language world model.
Audio
- Google release Gemini 3.5 Live Translate.
Science
- Anthropic: Making Claude a chemist (analyzing NMR spectra).
- OpenAI: A near-autonomous AI chemist improves a challenging reaction in medicinal chemistry. Paper: TEMPO Improves Generality and Decreases Oxidative Deboronation in Chan–Lam Couplings of Primary Sulfonamides.
- OpenAI: Using AI to help physicians diagnose rare genetic diseases affecting children. Paper: LLM-Assisted Reanalysis of Unsolved Rare Disease Genomes Increases Diagnostic Yield.
- Anthropic releases: Claude Science beta. Your research partner for rigorous science. The Claude Science app runs analyses, searches databases, and traces every step from data wrangling to publication, so you can spend time on science.
- 9 open, challenging math problems resolved using a harness and GPT-5.5 Pro and Claude Opus 4.8.
Hardware
- OpenAI describe their new custom chip design, Jalapeño, made with Broadcom: OpenAI and Broadcom unveil LLM-optimized inference chip.
- Etched reports on progress towards accelerated inference racks.
Robots