Mistral Medium 3.5 ships as the EU-friendly coding pick — 77.6% SWE-Bench Verified at open-weight Apache pricing

Mistral Medium 3.5, released April 29 and now widely available across cloud providers, hit 77.6% SWE-Bench Verified — putting it within striking distance of Qwen 3.5 and DeepSeek V4 on coding while shipping under Apache 2.0 from a Paris-based lab. For EU enterprises navigating data-residency-plus-IP-clarity procurement constraints, the model is the most defensible production-tier coding choice currently available.

The EU-friendliness is not a soft variable. Article 28 of the AI Act, the GDPR derogation framework for training data, and the upcoming Omnibus high-risk obligations all converge on a procurement preference for EU-headquartered, EU-trained, Apache-2.0-licensed models. Mistral Medium 3.5 is the only model that scores well on all three axes and meets the production capability bar.

The competitive implication for Cursor, Windsurf, and the in-IDE agent tier is direct. Enterprise procurement decks that previously routed to Claude or GPT family backends for coding tasks now have a third lane for EU customers. Cursor's model-agnostic routing was built for exactly this scenario; Anthropic's model-specific Claude Code product faces a harder pitch in EU enterprise sales.

Codersera — open-source LLM landscape 2026 → · Mistral AI news →