7 articles
Index of editorial tags and filtered coverage.
German search engine Ecosia announced this week that it is dropping models from France’s Mistral, which it says is “a year behind,” and switching to open-weight models such as Qwen, GLM and Kimi served by the European platform Melious, which it says has cut its costs in half.
The Financial Times reported on October 10, 2026, that Nvidia is in preliminary talks to acquire Reflection AI, the open-weight model startup in which it has already invested $800 million, or to deepen its stake with more capital or computing capacity.
Chinese company StepFun made its flagship Step 5 Preview model available on OpenRouter and Vercel AI Gateway on October 8, 2026, at $1 per million input tokens, and promises to publish its open weights on October 15.
Underdog, Conway Research’s on-device assistant, released Saluki 27B under the Apache 2.0 license, a 2-bit compression of Qwen3.8-27B that shrinks the model from 54 GB to 7.89 GB, improves at tool calling and loses 12 to 18 points on competition math.
EmbeddingGemma 2, built on Gemma 4 and released under the Apache 2.0 license, unifies text, code, images, video and audio in a single 768-dimension vector space, quadruples context to 8,192 tokens and gains 9.92 points on MTEB Code, with modular setups starting at 270 million parameters.
On October 6, Mistral AI opened API access to a preview of Mistral Large 4, a one-trillion-parameter multimodal model (49 billion active per task) that the company calls the most powerful open-weight system outside China; the weights will be released on October 27.
U.S. startup Reflection AI unveiled Beam, its first frontier open-weight model: a mixture-of-experts system with 501 billion total parameters and 23 billion active, aimed at coding, reasoning and agents, whose weights it promises to release this month under an Apache 2.0 license.