MiniMax has released M3.1 Flash, a mixture-of-experts model with 428 billion parameters, about 23 billion of them active, and a 1M-token context window. API pricing is $0.15 per million input tokens and $0.60 per million output tokens, and the weights are due on October 23.
JDN measured what OpenAI and Anthropic models actually cost on business tasks. The list prices published by the vendors turn out to be a poor guide to the final bill.
Google has released EmbeddingGemma 2, a compact open-weight multimodal embedding model based on Gemma 4 and licensed under Apache 2.0. It is designed to make semantic search and retrieval-augmented generation more efficient, including on devices.
Nvidia has released new open models aimed at companies that want to run their own models in their own data centers. The launch also pushes private AI agents that run on the device.
After months at the top of Arena, the crowd-sourced reference benchmark, Anthropic has been overtaken by Google and Meta. The ranking lists the 20 strongest models of the month.
Mistral has released Timestral, a family of time-series foundation models ranging from 5 million to 209 million parameters. The largest leads the GIFT-Eval, TIME and fev-bench benchmarks among zero-shot models and runs on a CPU.