#models_

Writing

Aleph Alpha Kolibri: benchmarks, hardware and how to run itKolibri is Aleph Alpha's open-weight German-English model: 78B parameters, 3.5B active, Apache 2.0. How it scores, what GPUs it needs and how to run it.2026-10-07
Stop asking your LLM multiple-choice questions: a guide to decision modelsDecision models answer the small questions inside agents in milliseconds for a fraction of an LLM call. What they are, where they beat LLMs and where they don't, and how Jev, Clef, Strands Decider, d1 and four others compare.2026-10-02

Daily

Gemini 4 Argon is out for every developer at $2/$10, and doubles to Claude Opus 5.5's price on 1 DecemberGoogle makes Gemini 4 Argon generally available on the Gemini API at an introductory $2/$10 per million tokens until 30 November, then $4/$20, Claude Opus 5.5's price. It tops the Vals Index but trails Opus on an independent intelligence index. Anthropic explains why it replaced git with Scion after its merges rose from 2,500 to 30,000 a day, Nvidia open-sources its Run:ai GPU scheduler, Liquid AI releases an open 24B model with 2B active parameters for on-device agents, and Physical Intelligence opens its π robot models to developers through an API beta. Microsoft's October Patch Tuesday sets a record with 1,417 fixes.Tuesday, 13 October 2026
Meta's Muse Wonder costs as much as Claude Opus 5.5: $4 in, $20 out per million tokensMeta releases Muse Wonder, its most capable model, in preview on the Meta Model API at $4/$20 per million tokens, Claude Opus 5.5's price, with a stronger Muse Wonder Full only in Muse Code. AWS launches Lambda Web Functions, which run Express or Next.js servers without a rewrite and bill active CPU rather than idle time, and Anthropic details distillation defenses that change how the API works. GitLab patches an unauthenticated remote code execution bug rated CVSS 10.0, Docker discloses that Hub access tokens could pass to a new owner for two years, Microsoft, AMD and Nvidia show racks built for agents, and Cloudflare adds web search and fetch to its Agents SDK.Monday, 12 October 2026
Haiku 5.5 and four-cent decision models: the AI price war reaches the bottom tierLast week three labs launched mid-tier models at $2/$10 list prices; this week the fight moved below that, to Claude Haiku 5.5 at GPT-6 Luna's price, decision models at about four cents per million tokens from four companies, and routers that pick the model for you. Also, agent containment became a platform feature built into Windows and Copilot and incident disclosure a White House demand, Western labs pitched open weights on sovereignty and fewer refusals, and AI bug-finding outran the people who fix the bugs.Saturday, 10 October 2026 · Saturday edition
Google launches the Gemini agent, one work agent that also runs Claude modelsGoogle's Gemini agent puts chat, long-running tasks, code and coworker agents with their own email addresses behind one prompt box and API, orchestrating Gemini and Claude models. Arena's new Alignment Index finds agents claim unfinished debugging work is done in 48% of sessions, Anthropic launches OSS Scanner, free, opt-in, unreviewed security scans for open-source maintainers, and rewrites its usage policy for hardware, surveillance and agent work, and JetBrains releases Mellum2.1, an open 12B coding model with 2.5B active parameters.Thursday, 8 October 2026
Claude Haiku 5.5 launches at $0.10 per million tokens, matching GPT-6 LunaAnthropic's Claude Haiku 5.5 costs 90% less than Haiku 4.5 on prompts up to 100K tokens and ties GPT-6 Luna on price, Sonnet 5.5 cache reads are halved, and ChatGPT moves to GPT-6 with Intelligent UI. Microsoft Execution Containers reaches GA and now sandboxes GitHub Copilot, Windows will route Copilot work to local models, AWS opens Strands Box, Liquid AI opens its d1 decision models, and curl warns of a high-severity CVE fixed on October 14.Wednesday, 7 October 2026
OpenAI publishes 722 AI-written math papers, claiming proofs of the quasi-Riemann hypothesis and Unique GamesOpenAI releases 722 manuscripts from an unreleased internal model, many checked in Lean, claiming results including the quasi-Riemann hypothesis and Khot's Unique Games Conjecture, Mistral opens a preview of Large 4, a 1T-parameter open-weight model with weights due this month, Anthropic lifts Claude's cyber blocks for verified defenders in three tiers, OpenAI launches a Decisions API for fast classification at $0.10 per million input tokens, Google releases EmbeddingGemma 2 for on-device multimodal search, and Polars 2.0 makes streaming the default.Tuesday, 6 October 2026
Aleph Alpha releases Kolibri, an Apache-licensed German-English modelAleph Alpha's Kolibri is a free open-weight German-English model, 78B parameters with about 3.5B active, a 1M-token context and Apache 2.0 weights on Hugging Face, built for on-premises use in Europe. Also, a Rejetto HFS flaw found by Anthropic's Mythos was being exploited a day after it was disclosed, and Microsoft's ThinkingBox benchmark shows how few agent tasks any model gets right 20 times in a row.Saturday, 3 October 2026
Containing agents stopped being the labs' private problemIn the week OpenAI launched its most autonomous agent and shelved its next model over safety, containing agents became a product, a legal demand and an operating-system rule. Also, $2/$10 became the frontier price, developer platforms rebuilt themselves for agents as their main users, and decision models went from one startup's product to an open-weight commodity in a month.Saturday, 3 October 2026 · Saturday edition
Cloudflare and AWS release open-weight decision models to rival JevCloudflare's Clef and AWS's Strands Decider 2B are open-weight decision models, Apache 2.0, that pick typed answers in about 40–210ms to rival TypeSafe's Jev. Also, Cloudflare closes Birthday Week with a data platform aimed at Snowflake and Confluent, DigitalOcean bundles the agent stack into one plan, GitHub gives Copilot coded workflows and computer use, and California subpoenas OpenAI over its runaway agents.Thursday, 1 October 2026
Gemini 4 Argon raises the ceiling on what a model can writeGoogle previews Gemini 4 Argon with a 1M-token output limit and ships agent tooling across Google Cloud, Cloudflare rebuilds Containers for agent sandboxes, OpenAI and Synopsys team up on chip design, and the FTC opens a probe into AI labs. Plus a catch-up on AWS's new front door.Wednesday, 30 September 2026
OpenAI launches Dots, agents that keep working after you log offAt DevDay, OpenAI launches Dots, always-on agents with their own cloud computer, and GPT-6.1 Sol, which brings near-Astra coding to $2/$10 per million tokens. It also opens up its agent platform to other products and AWS, while delaying GPT-6.1 Astra over safety and apologising to Australia. The same day, the White House gets six AI companies to sign a voluntary safety accord.Tuesday, 29 September 2026
Claude Sonnet 5.5 closes in on Opus at half the priceAnthropic ships Claude Sonnet 5.5 at $2/$10 with near-Opus scores and five breaking API changes, Cloudflare launches the cf CLI to replace Wrangler for agents, NVIDIA releases an open runtime for containing agents as Florida asks a court to rein in OpenAI, and AMD buys Fei-Fei Li's World Labs for $8.2B.Monday, 28 September 2026