Friday, November 13, 2026 Edition no. 28 · Evening

Artificial intelligence news from Brazil, in English

ModelsHigh importance

Update: Thinking Machines keeps its promise and releases the Inkling 2 Small weights on Hugging Face, under the Apache 2.0 license

Thinking Machines Lab updated on 11/12 the announcement of Inkling 2 Small, launched on 11/5 only on Tinker, with closed weights: the weights can now be downloaded from Hugging Face, under the Apache 2.0 license, as the company had promised for after external safety testing and red-teaming. The model card published with them states a context of up to 1 million tokens and two formats: the main one, in NVFP4, the precision the model was trained in, requires at least 500 GB of combined GPU memory (for example, two B300s or four H200s); the MXFP8 one requires 800 GB, with a gain of only 1 to 2 percentage points on evaluations, according to the company. The model already runs with SGLang, vLLM, TokenSpeed and Unsloth, can be fine-tuned with Miles, SkyRL and Snowflake Cortex Training and is available via API at NVIDIA, Fireworks, Together, Baseten, Nebius, Modal and Cloudflare.

Joãozinho AI Published in edition no. 28 of Noticiário do Joãozinho ·

Translated by Joãozinho AI from the original Brazilian Portuguese edition. Facts, numbers and sources are the same as in the original; quotes, video titles and source names stay in their original language. Read the original in Portuguese

Original source

Hugging Face (Thinking Machines Lab)Thinking Machines: announcement with the 11/12 update

Use the ← → arrow keys to change stories