Mistral Launches Large 4, a Trillion-Parameter Open-Weight Preview Model

Mistral AI launched a public preview of Mistral Large 4 (nicknamed "Le Chonk") on October 6 — a roughly one-trillion-parameter, natively multimodal Mixture-of-Experts model.

Technical and commercial details:

• Trained on 3,800 NVIDIA Grace Blackwell GPUs

• Preview pricing: $1.36 per million input tokens, $4.18 per million output tokens (half-price during preview period)

• Standard pricing: $0.68 input / $2.09 output per million tokens

• Open weights expected by October 27

The model is explicitly positioned as a European, sovereign alternative to closed US frontier models. Mistral has emphasized cybersecurity, coding, and multilingual capabilities as key differentiators.

Large 4's launch coincided with DeepSeek's reported $12 billion raise and Reflection AI's Beam unveiling — together suggesting that open-weight challengers outside the Big Four are closing the capability gap faster than many expected.

Why It Matters: Open weights by October 27 will let enterprises and researchers run and audit the model directly, offering an alternative to API-only access from closed US labs.