Mistral Medium 3.5 (mistral.ai)

Article summary

Mistral AI has released Mistral Medium 3.5, a new 128B open-weight model optimized for long-horizon coding and productivity work. The model is available in public preview and powers remote coding agents and Work mode in Le Chat on the Pro, Team, and Enterprise plans. Mistral Medium 3.5 is priced at $1.5 per million input tokens and $7.5 per million output tokens. The model is designed to run on as few as four GPUs and has strong real-world performance.

Main themes

• AI model releases
• Cloud-based coding agents
• Long-horizon productivity
• Model pricing and availability
• Global AI competition

What commenters say

• The new Mistral Medium 3.5 model is not significantly better than other available models, and its performance is not impressive compared to the current state-of-the-art models.
• The model's pricing is expensive considering its size and performance, and it may not be competitive in the market.
• The fact that Mistral is a non-US and non-Chinese model is a significant factor in its appeal, especially for European customers who may prefer to use models from other regions.
• The global AI competition is dominated by US and Chinese companies, and other countries such as Europe, Japan, and Korea are not producing competitive models.
• The development of AI models is heavily influenced by funding and investment, and companies may need to move to the US to access more funding and resources.
• Some commenters argue that the comparison to other models is unfair, and that Mistral Medium 3.5 has its own strengths and advantages.
• The model's performance on specific benchmarks, such as SWE-Bench Verified, is not impressive, and its claims of exceeding other open models are questionable.
• The availability of models from other countries, such as Korea and India, is limited, and they are not yet competitive with models from the US and China.