Key Findings
Mistral Large 4 (nicknamed “Le Chonk”) has been released as an open-weights model following three additional weeks of training after its October 6 API preview. The primary improvements are concentrated in coding efficiency; specifically, the model achieved a 41% score on Terminal-Bench 4.0, up from 28.3% in the preview version. Mistral also reports a 30% reduction in output tokens and a 47% reduction in average tokens per task compared to GLM-5.3 on the Coding Agent Index, where it scored 48%.
Technical Details
The model is a Mixture of Experts (MoE) architecture with approximately 1.05 trillion total parameters. Users can configure reasoning depth across three levels: “low,” “medium,” and “high” (the default). In terms of API pricing, Mistral Large 4 is priced at $0.70 per million input tokens, $0.07 per million cached input tokens, and $2.10 per million output tokens. Artificial Analysis reports an output speed of 161.3 tokens per second for the “high” reasoning setting.
Comparative Performance
Despite coding gains, general intelligence evaluations show a gap with Chinese competitors. According to the Intelligence Index v5, Mistral Large 4 (high setting) scored 26 overall, while GLM-5.3 and DeepSeek V4.1 Flash 1017 both scored 32. Regarding API costs per task, Mistral Large 4 is more affordable than GLM-5.3 but more expensive than DeepSeek V4.1 Flash 1017.
Significance and Outlook
The release allows organizations to deploy the model in their own computing environments, though this requires securing GPUs and operational staff. Because coding performance does not correlate linearly with overall intelligence scores, adopters must determine if the specific gains in terminal operations and token efficiency justify the transition. The discrepancy between different benchmark sources, such as FrontierCode 2.0 results from Mistral versus Artificial Analysis, suggests that users should prioritize testing tasks close to actual use cases over aggregate scores.
Source: https://xenospectrum.com/en/mistral-large-four-open-weights/
Get our weekly technology intelligence — free
Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.
Subscribe Free — Weekly Tech Intelligence
By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.
- Your email and selected fields are used only to deliver the newsletter.
- We never share your information with third parties.
- You can unsubscribe anytime via the link in each email.
See our Privacy Policy for details.
Takes about a minute · Unsubscribe anytime
