Mistral AI Updates: October 31, 2026

1. Mistral Prices Large 4 at $0.70 Input and $2.10 Output per Million Tokens and Previews a Cybersecurity Variant for Design Partners

Mistral AI published its own post on the Mistral Large 4 (“Le Chonk”) weight release, adding details the Hugging Face launch post did not include. API pricing is $0.70 per million input tokens, $0.07 for cached input and $2.10 per million output tokens. Hosted access covers Mistral Studio, the Mistral API, OpenRouter, NVIDIA-hosted endpoints and Microsoft Foundry, which Mistral calls the first cloud provider to offer the model; self-hosting options are the Hugging Face weights with vLLM or SGLang, an NVIDIA NIM container and TensorRT-LLM, and customization runs through Mistral Forge or NeMo AutoModel. A specialized cybersecurity variant is in preview only, for design partners such as Orange Cyberdefense under vetting and contractual safeguards, and Mistral says it observed some uplift potential in live traffic and is adjusting its guardrails. Mistral-reported figures: 19.7% on Cognition’s private FrontierCode 2.0, 48% on the Artificial Analysis Coding Agent Index (tied with GLM-5.3 while using 47% fewer tokens per task), 41% on Terminal-Bench 4.0 (up from 28% in the October preview, with 30% fewer output tokens), 51% on the Artificial Analysis Cyber Index, a 96.6% average refusal rate on JailbreakBench, StrongREJECT and AgentHarm, 97.1% attack resistance on the B3 agent security benchmark, 73.3% on FinWorkBench and 84.5% on BrowseComp. Reasoning effort is low, medium or high (the default), hardware requirements are not stated, and a technical report is promised. Source