Mistral AI has released Mistral Large 4, a state-of-the-art open-weight multimodal model featuring a granular Mixture-of-Experts architecture. The model contains 49 billion active parameters and 1.05 trillion total parameters, along with a 1.6 billion parameter vision encoder. It supports a 1 million token context window and is designed for general-purpose tasks across text and vision modalities.
Pricing is set at $1.36 per million input tokens and $0.68 per million cached input tokens, with output tokens priced at $0.14 per million and cached output at $0.07 per million. The model includes built-in support for structured outputs, function calling, document question answering, prefix chat completions, batching, and agent-based conversations. Mistral Large 4 is positioned as a high-performance, cost-efficient option for developers and enterprises seeking advanced multimodal capabilities with open weights.
Source: Hacker News · Summarized by HeadlinesBriefing