French AI Company Mistral AI Launches Open Source Model Mistral Medium 3.5, Market Response Tepid

On May 1, French AI startup Mistral AI released its next-generation open-source model, Mistral Medium 3.5, on April 29. This model features a 128 billion parameter dense architecture and is accompanied by the launch of the remote coding agent tool Mistral Vibe CLI and the multi-step AI Agent functionality of Le Chat. Official data shows that Mistral Medium 3.5 scored 77.6% on the SWE-Bench Verified programming benchmark and supports unified reasoning and coding capabilities. However, its API pricing, which reaches $1.50 per million input tokens and $7.50 per output token, is significantly higher than that of Chinese open-source models, leading to widespread skepticism in the community. Currently, the top positions in the open-source model rankings are dominated by Chinese models such as Alibaba’s Qwen, Zhipu GLM, and Xiaomi’s MiMo-V2. Notably, Qwen 3.6 has only 27 billion parameters but has achieved a score of 72.4% on SWE-Bench and is completely open-source and free. Nevertheless, some developers believe that Mistral’s greatest value lies in its European identity, being ‘non-American, non-Chinese.’ Due to compliance with GDPR and local deployment requirements in Europe, its models have been utilized for privatized deployment by large European institutions, including HSBC. Market perspectives suggest that Mistral’s competitiveness stems more from geopolitical factors and corporate compliance rather than pure model performance.