Mistral Medium 3.5 Launch Creates Disappointment

Mistral Medium 3.5 Launched

Mistral AI launched the Mistral Medium 3.5 model on April 29, disappointing the AI sector. The Paris-based company introduced a dense model with 128 billion parameters, remote coding agents, and a new work mode in the Le Chat interface. Users were impressed by cloud-based coding sessions capable of sending pull requests directly to GitHub, but the overall reaction remained cool. The model drew engineering attention by combining three separate architectures (decoder-only, encoder-decoder, and MoE-like) into a single weight set. Pricing sparked debate: $1.50 for input per million tokens, $7.50 for output. This was found expensive compared to competitors like OpenAI's GPT-4o-mini.

Performance and Competitor Comparison

Medium 3.5 achieved 77.6% success in the SWE-Bench Verified coding test and reached 91.4% in τ³-Telecom, which measures agentic tool usage. Alibaba's Qwen 3.6 model, with 27 billion parameters, competes in similar benchmarks while being freely downloadable. In open-source leaderboards, Chinese-origin models like Zhipu AI's GLM and Xiaomi's MiMo-V2 dominate the top spots. Mistral Vibe CLI offers parallel coding sessions, while Work Mode handles multi-step tasks like email classification and research synthesis. These agents hold potential for blockchain developers in smart contract coding; for example, they can be applied in projects like ALT detailed analysis.

Geopolitical Dynamics and Reactions

Reactions are mixed: figures like Pedro Domingos and Youssof Altoukhi criticized the pricing and performance, while some developers value the European-origin alternative. Internet users emphasized that the model is expensive and mediocre compared to competitors. The launch reveals Western AI struggling against Chinese dominance; Mistral is the sole European representative. GDPR-compliant institutions and banks with sensitive data prefer Mistral to avoid US or Chinese infrastructure. Companies like HSBC host it on their own servers. For AI's crypto integration, ALT futures trading should be monitored; regulations shape the price-performance competition. In the sector, Europe maintains its position through lobbying power.