Mistral Codestral 2 Released
Narrative
Specialized coding model. 32K context. Fill-in-middle support. 85+ programming languages. €1/M tokens.
Reality
Benchmarks competitive: 92.8% HumanEval, 89.3% MBPP. Fill-in-middle excellent for IDE integration. Pricing extremely aggressive. European developers adopted rapidly.
Implication
Established specialized model viability. Coding became commoditized. European sovereignty angle resonated. Price pressure on OpenAI Codex intensified.