Round-up
Oct 7, 2026
Mistral Large 4 enters public preview
Mistral introduced a public API preview of Large 4, its trillion-parameter multimodal model, with weights promised for later this month rather than available now. OpenAI published mathematical manuscripts and proof artefacts from an unreleased frontier model; verification varies across the collection, so the release is not blanket confirmation of every claimed result. Anthropic expanded its Cyber Verification Program with tiered access to advanced Claude models for vetted security work, while Microsoft examined the practical burden of AI-driven vulnerability discovery.