Models
Frontier and open models: what shipped, what it can do, and how the claims hold up.
-
Xiaomi publishes MiMo-V2.6 and the recipe for scaling its reinforcement learning
Xiaomi’s new MiMo-V2.6 report is a reinforcement learning scaling recipe, with open weights under an MIT licence and, unusually, no benchmark headline…
-
Sakana AI’s Japanese-tuned model goes into a clinical evidence search tool
Sakana Namazu is now behind Evidence Finder, a doctor-facing literature search tool, with a 96.4% medical licensing exam score and a very…
-
Abu Dhabi’s TII releases a speech model aimed at Emirati Arabic
TII’s Falcon-ASR reports the best published error rate on an Arabic leaderboard, while its Emirati dialect result rests on the institute’s own…
-
LightOn ships three OCR models and picks apart its own benchmarks
LightOn has released OCR models at three sizes and spends much of the announcement explaining why the edit-distance scores it reports are…
-
Microsoft brings its MAI Code model onto the PC, and lets GitHub Copilot route work to it
Microsoft is shrinking its MAI Code 1.1 Flash coding model to run on Windows PCs, and GitHub Copilot will start routing some…
-
Cloudflare releases its first open-weight models, and they do not write text
Clef and Clef-flash answer typed questions with a probability for each allowed option instead of generating prose, and the weights are on…
-
Aleph Alpha releases Kolibri, a German-first open-weight model, and writes down its data screening
Kolibri is a 78 billion parameter mixture-of-experts model built for German, released under Apache 2.0, with a technical report that documents how…
-
Nvidia reports olympiad gold twice over, under two very different conditions
Nvidia says Nemotron 3 models hit gold standard at this year’s informatics and mathematics olympiads, but only the maths proofs were graded…
-
Anthropic cuts small-model pricing by 90% with Claude Haiku 5.5
Claude Haiku 5.5 costs $0.10 per million input tokens on prompts up to 100,000 tokens, which Anthropic says is 90% below Haiku…
-
OpenAI puts GPT-6 in front of everyone, and publishes where it got worse
OpenAI is rolling GPT-6 out to every ChatGPT tier worldwide, and its accompanying safety update names the evaluations on which the new…