Model releases
Meta launches Muse Wonder, an agentic model with 1M context, on its API
Muse Wonder adds Active Memory and Active Focus for long agent tasks. It is on the Meta Model API at $4/$20 per…
Oct 12, 20262 min read
Model releases
Anthropic releases Claude Haiku 5.5 at $0.10 per million input tokens
Claude Haiku 5.5 adds an effort setting, beats Haiku 4.5 on every listed benchmark, and costs up to 90% less for prompts…
Oct 7, 20262 min read
Model releases
Mistral Large 4 enters public preview; open weights due end of October
Mistral's ~1T-parameter MoE model (52B active, 1M context) is live in preview at $1.36/$4.18 per 1M tokens. Weights are promised by month's…
Oct 6, 20262 min read
Model releases
Google releases EmbeddingGemma 2, an open on-device multimodal embedder
EmbeddingGemma 2 (740M, Apache 2.0) embeds text, code, images, video and audio in one space, with 8K context and about 191MB RAM…
Oct 6, 20262 min read