Mistral AI launched its latest model, Mistral Medium 3.5, on April 29, introducing a dense 128-billion-parameter architecture and a suite of new features. However, the Paris-based lab’s efforts have been met with a wave of lukewarm online responses. The release comprises three key components: the model itself, the Mistral Vibe CLI for remote coding agents that facilitate cloud-based coding sessions, and the Work Mode in Le Chat, Mistral’s ChatGPT-like consumer interface, capable of handling multi-step tasks such as email management and research synthesis.
Despite the ambitious scope, initial benchmark results present a mixed picture. Mistral Medium 3.5 achieves a score of 77.6% on the SWE-Bench Verified coding benchmark, which assesses the model’s ability to address live GitHub issues by generating functional patches. It fares better on the τ³-Telecom benchmark, scoring 91.4%, which evaluates agentic tool use within specialized settings. Notably, Mistral has consolidated three previous models—Medium 3.1, Magistral, and Devstral 2—into a single set of weights, allowing for configurable reasoning effort per request, marking a significant engineering achievement.
However, the pricing structure poses concerns. Mistral charges $1.50 per million input tokens and $7.50 per million output tokens. In contrast, Alibaba’s Qwen 3.6, which contains only 27 billion parameters, achieves a score of 72.4% on the same SWE-Bench benchmark while being available under the Apache 2.0 license, allowing users to download and run it for free. Generally, parameters correlate with an AI model’s learning capacity; hence, a higher parameter count typically signifies a broader knowledge base.
The competitive landscape becomes clearer when examining open-source leaderboards. Leading contenders such as Alibaba’s Qwen, GLM from China’s Zhipu AI, and MiMo-V2 from Xiaomi dominate the top spots, proving to be cheaper and more effective than Mistral’s latest offering. Following its release, Medium 3.5 has not yet made a significant impact on major independent leaderboards, as third-party evaluations remain outstanding.
The only encouraging aspect, some observers contend, is that Mistral is currently the sole non-Chinese player with a notable presence in the open-source conversation. Machine learning professor Pedro Domingos from the University of Washington expressed skepticism regarding Mistral’s performance, stating, “Regular AI companies brag about how much better their model is on benchmarks. Only Mistral brags about how much worse its one is.” He further questioned the implications of European representation in the AI race, stating, “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.”
Responding to the competitive pricing, Youssof Altoukhi, founder of Yoyo Studios, pointed out that Qwen 3.6 is 4.7 times smaller than Medium 3.5 while achieving comparable coding scores. He remarked, “If it wasn’t for their political skill they would have been bankrupt by now.” Not all feedback was negative, however. AI developer Michal Langmajer expressed a more nuanced view, acknowledging, “I’m genuinely glad there’s still a non-US, non-Chinese lab trying to build frontier LLMs but boy we have to level up the game in Europe.” He noted that Mistral’s flagship model ranks poorly on benchmarks while carrying significantly higher costs than many rivals.
Some developers argued that open weights present a durability advantage over pure leaderboard performance, suggesting that a model that can be downloaded, fine-tuned, and self-hosted does not necessarily need to dominate rankings to remain relevant. Others pointed to Mistral’s enterprise deployments throughout Europe as evidence that its advantages extend beyond technical capabilities.
Mistral’s positioning underscores a broader geopolitical narrative. With stringent regulations like the GDPR in Europe, companies handling sensitive customer data are often wary of routing AI workloads through Chinese infrastructure. This has led to significant contracts, such as HSBC’s multi-year agreement with Mistral to enable self-hosting of models on its infrastructure. While Mistral may not lead in coding performance or cost, its appeal as an EU-based, open-weight lab provides a legally compliant option for European enterprises, thus reinforcing its value in procurement decisions.
See also
Germany”s National Team Prepares for World Cup Qualifiers with Disco Atmosphere
95% of AI Projects Fail in Companies According to MIT
AI in Food & Beverages Market to Surge from $11.08B to $263.80B by 2032
Satya Nadella Supports OpenAI’s $100B Revenue Goal, Highlights AI Funding Needs
Wall Street Recovers from Early Loss as Nvidia Surges 1.8% Amid Market Volatility
