Mistral AI has released the weights for its 1-trillion-parameter model, Mistral Large 4, under a modified MIT license. The company claims this model leads in open-weight coding benchmarks and demonstrates strong performance in cybersecurity, refusing 96.6% of malicious cyber requests and scoring 51% on a defensive cyber index for flaw detection and patching. AI
IMPACT This release provides a powerful open-weight model for coding and cybersecurity tasks, potentially accelerating research and development in these areas.
RANK_REASON Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
Read on Mastodon — mastodon.social →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →