On April 29, Mistral AI introduced the Mistral Medium 3.5 open-source model, which boasts an impressive size of 128 billion parameters. This release includes not only the model but also additional tools such as remote coding agents via Mistral Vibe CLI and Work Mode in Le Chat. The Medium 3.5 model achieved benchmark scores of 77.6% on SWE-Bench Verified and 91.4% on τ³-Telecom, highlighting its capabilities in specific technical assessments.
Mistral Medium 3.5 represents an integration of previous models Medium 3.1, Magistral, and Devstral 2 into a consolidated framework featuring a single set of weights. This configuration allows for adjustable reasoning effort per request, providing flexibility for different computational needs. The model is characterized by its extensive 128-billion-parameter architecture, highlighting its capability to manage complex computational tasks. Additionally, the release includes access to advanced remote coding agents through the Mistral Vibe CLI, enhancing utility for developers. Users also benefit from the “Work Mode” feature available in Le Chat, which facilitates an interactive environment for enhanced model interaction.
The Mistral Medium 3.5 model is priced at $1.50 per million input tokens and $7.50 per million output tokens, reflecting a structured pricing model based on usage. In terms of performance, the Mistral Medium 3.5 achieves a score of 77.6% on the SWE-Bench Verified and 91.4% on the τ³-Telecom benchmarks, suggesting robust capabilities in certain technical areas.
In comparison, Alibaba’s Qwen 3.6 model, with its 27 billion parameters, scores slightly lower at 72.4% on the SWE-Bench Verified benchmark. Unlike the Mistral model, Qwen 3.6 is released under an Apache 2.0 license, allowing it to be downloaded and run for free, contributing to its open-source accessibility.
Leading open-source AI models identified in available reporting include Alibaba’s Qwen, GLM from Zhipu AI, and MiMo-V2 from Xiaomi. These models are named as top open-source leaders in the current ecosystem. Mistral Medium 3.5 enters this landscape alongside those contenders. The availability of multiple open-source offerings is noted in the reporting.
Mistral Medium 3.5 has not yet ranked on major independent leaderboards, and third-party evaluations are pending. “Regular AI companies brag about how much better their model is on benchmarks. Only Mistral brags about how much worse its one is.” “I don’t know what’s worse, for Europe to not be in the AI race or for it to be represented by a laughingstock like Mistral.”
Mistral AI released the Mistral Medium 3.5 open-source model on April 29, a 128-billion-parameter architecture that was issued together with remote coding agents via Mistral Vibe CLI and an interactive Work Mode in Le Chat. The release is offered under a usage pricing of $1.50 per million input tokens and $7.50 per million output tokens, and the model has not yet appeared on major independent leaderboards while competing in an ecosystem that includes Alibaba’s Qwen, Zhipu AI’s GLM, and Xiaomi’s MiMo-V2.
