Mistral Large 4 ("Le Chonk"): 1T-Parameter Open-Weights Multimodal Model
Melvin Vivas · X video post · 2026-10-06 · 0:17 · 988 views · Open on X
Topics: LLM Fundamentals, Industry Trends & Job Market · Level: beginner
Summary
This short post shares Mistral AI's announcement of Mistral Large 4, which Mistral nicknames "Le Chonk". Mistral says it is an open-weights, natively multimodal model with 1T total parameters and 49B active per token, which points to a mixture-of-experts design. Mistral calls it the best open-weights model from the US or Europe on aggregated benchmarks. The creator says he has never used Mistral and asks what it is good at. The video itself has no real content.
Key points
- Mistral Large 4 is a new model from Mistral AI, nicknamed "Le Chonk".
- It has 1 trillion total parameters, but only 49B are active per token. That ratio suggests a sparse mixture-of-experts (MoE) design.
- Mistral says it is natively multimodal and released with open weights.
- Mistral claims it is the best open-weights model from the US or Europe on aggregated benchmarks.
- Mistral claims state-of-the-art results on high-stakes workloads: cyber defense, manufacturing and finance.
- Total vs. active parameters matters when choosing a model: active parameters drive the compute cost per token, while total parameters drive how much memory you need to host it.
Resources mentioned
- Mistral Large 4 · tool · docs.mistral.ai · free
Mistral AI's open-weights, natively multimodal model with 1T total and 49B active parameters. - Mistral AI · website · mistral.ai · check price
French AI lab that publishes open-weights and commercial LLMs, plus an API and the Le Chat assistant.
Also in: Real-Time Speech Transcription in the Browser with Voxtral and WebGPU (Melvin Vivas on X · notes)
More in LLM Fundamentals
- Claude Haiku 5.5 in Devin: beats Sonnet 5 at about 1/8 the cost
- Liquid AI's open d1 decision models: small multimodal models that run locally
- Comparing Opus 5.5, GPT-6 Astra and Grok 4.7 on OpenRouter
- Liquid AI's d1 Decision Model Now Accepts Images
- Run Local LLMs on the Computer You Already Have: A Beginner's Path
- Use Nvidia Nemotron 3.5 Lightning on OpenRouter for Cheap Worker Tasks