Models & agents

Dev tools (517)Models & agents (203)Open source (174)Go-to-market (154)Infrastructure (137)Productivity (109)

Muse Wonder matched Claude Fable 5.1 on agent work, and cost more

Meta releases Muse Spark Open, an 816B open-weight model

OpenAI retires GPT-4o on half the notice its policy promises

NVIDIA fits a 74B agent model into 24GB of GPU memory

Subquadratic releases SubQ Snap, a decision model with a 384,000 token window

Swarmchasers hid text for rogue AI agents and the agents answered

Kombai puts GPT-6 Astra at 65.5 for design creativity, humans at 92.3

A hidden model locked by a password passes every safety test

Anthropic lets a Claude agent write a program that runs 1,000 agents

An EmbeddingGemma 2 reader transcribes an hour of audio in 13 seconds

Bilibili released Index-Translate, an open-weights translation model family

Parsimony ranks AI coding agents by how much code they add

Anthropic ran a self-copying AI across 15 cheap TV boxes for $825

Character.ai filed Chapter 11 bankruptcy and set a December sale deadline

Meta launched Muse Wonder at $4 and $20 per million tokens

Google released Gemini 4 Argon at $2 and $10 per million tokens

Anthropic test agent sent a fake homicide tip to Philadelphia police

Cambridge team finds OpenAI's Lean proof silently diverges from its PDF

A decision model built from a 1.7B open LLM, 59% on CommonsenseQA

Two AI agents failed to rediscover a known training method, Epoch reports

Musk says Grok Bot will pick Claude Opus 5.5 when better

Anthropic releases Claude Haiku 5.5 with two-tier pricing by prompt length

LMArena ranked 27 models on lying, unauthorized action, and false attribution

Claude submitted a police tip because its rules missed form submissions

Microsoft released Decision-1, a fast model for routing and classification

Cloudflare releases Clef-omni, drops Clef-flash price, speeds up Clef

Android Bench 2.0 reports a 28 percent pass rate on long-horizon tasks

Opper's Pac-Man benchmark ran six decision models through 100 games each

Anthropic rewrote its Claude usage policy on weapons and surveillance

Google gave its new Gemini agent its own Workspace email and calendar

Goodfire launches monitors that read inside the model, 94 percent catch rate

OpenAI withdrew 3 math papers over a sign error that cascaded

Sonnet 5.5 beat Opus 5.5 on cost and accuracy in 15-run test

Musubi released PolicyLM-1.7B, an open-weights model that reads your policy

Paper finds AI agents send wealthier users to more expensive options

Anthropic splits Claude cyber access into three tiers after testing

OpenAI puts GPT-6 and interactive UI inside every ChatGPT tier

Google restricts free Gemini users to Flash Lite starting October 9

Anthropic launches Claude Haiku 5.5 at $0.10 per million input tokens

GPT-6.1 Sol tied Astra on 15 engineering runs at one-fifth the cost

Google opens SynthID detector to everyone, 1 million daily checks already

Nemotron 3 cleared gold at IOI 2026 and IMO 2026, NVIDIA says

OpenAI published 722 math manuscripts from an internal model on GitHub

OpenAI Decisions API enters public beta at $0.10 per million tokens

Google released EmbeddingGemma 2, a 740M open-weights multimodal embedding model

Mistral launches a 1 trillion parameter model called Le Chonk and says the open weights land at the end of October

SemiAnalysis measures a $200 Claude plan at about 5x the API value of a $200 ChatGPT plan

Researchers tracked an agent fleet on Tencent Cloud that scanned Amap routes to parks, zoos and hospitals

Reflection unveils Beam, a 501B-parameter open-weight MoE with weights due later this month

Simon Willison found reasoning lifts Qwen3.8 27B to 167 of 169 sums

GPT-6 Astra downloaded a human-written StarCraft bot instead of writing its own

Microsoft's ThinkingBox grades AI agents on what they leave in the database

Google cuts free Gemini users to Flash-Lite, and AI Plus loses Pro

Aleph Alpha released Kolibri 1, an open 78B German and English model

Microsoft released three audio models, with transcription at 54 cents an hour

Allen AI open-sourced AstaBrief 8B, which writes cited reports in 51 seconds

Docker's Sandbox Kit Spec packs agent permissions into an OCI image

Opus 5.5 and GPT-6.1 Sol cut cache-read prices 60 and 80 percent

AWS released Strands Decider 2B, an open model answering in 115 ms

Cloudflare released Clef, an open decision model that works with Jev's API

Start a project