Datasette Agent 0.5a0, Fine-Tuning Updates, and More
This week the common thread is tighter feedback loops. Tools are moving the model, the data, and the...
Tag archive
This week the common thread is tighter feedback loops. Tools are moving the model, the data, and the...

On 4 November 2026, Moonshot AI completed the promise it made at launch: the weights of Kimi K3.1 Pro...

LongCat-2.5 open weights ship under MIT, ahead of Kimi K3.1 Pro and ERNIE 5.5 Flash Base on licence terms. Rent the API first, self-host later.

SplatAnything beats Mirror2-GS on most of seven 3D benchmarks and its open weights are out, but the licence excludes the EU, UK and South Korea.

PLaMo 3 Translate open weights from Preferred Networks run offline in 53 languages. We compare it with TranslateGemma and Hy-MT2 and pick a winner.

Moonshot's Kimi K3.1 Pro scores 65.7% on Terminal-Bench 4.0, a point behind Claude Opus 5.5. Its weights are due by November 4, with cyber data removed.

Muse Spark Open is on Hugging Face right now, weights and all, tuned for Apple silicon. The catch is in the system requirements: two Mac Studios lashed together, or four of them at roughly $38,000.

A decision model picks from fixed answers and returns a probability for each, instead of writing text. Hosted prices start at $0.02 per million tokens.

Meta released Muse Spark Open, an open-weight model with 816B parameters, 42B active and a 1M-token context, on Hugging Face and three cloud platforms.

Mistral's new Mistral Large 4 — officially nicknamed le Chonk — is a 1-trillion-parameter open-weight model it says beats anything out of the US or Europe. Weights land at the end of October. The underdog just got heavy.
Probabilità al posto di testo, latenze più basse e supporto immagini: una nuova strada per...

Google DeepMind's EmbeddingGemma 2 maps text, code, images, video and audio into one vector space with 740M parameters. Text-only use needs about 191 MB of RAM.