W3BStation
Markets
BTC $96,420 +2.34% ETH $3,280 +1.82% SOL $185.40 -0.92% BNB $642.50 +0.45% XRP $2.18 +3.12% DOGE $0.082 -1.50% ADA $1.05 +0.80% AVAX $42.10 +1.15%
BTC $96,420 +2.34% ETH $3,280 +1.82% SOL $185.40 -0.92% BNB $642.50 +0.45% XRP $2.18 +3.12% DOGE $0.082 -1.50% ADA $1.05 +0.80% AVAX $42.10 +1.15%
10/07/2026

Mistral AI Launches 1T-Parameter 'Le Chonk' Model, Outperforming GPT-6 Astra on Finance Benchmarks

What happened: Paris-based Mistral AI has released a preview of its Large 4 model, nicknamed 'Le Chonk.

Mistral AI Launches 1T-Parameter 'Le Chonk' Model, Outperforming GPT-6 Astra on Finance Benchmarks

What happened: Paris-based Mistral AI has released a preview of its Large 4 model, nicknamed 'Le Chonk.' This 1.05 trillion-parameter mixture-of-experts model, with 49 billion active parameters per inference, is natively multimodal and offers a 1 million-token context window. On the Finance Agent v2 benchmark, it scored 54.7, surpassing GPT-6 Astra (53.5) but trailing Claude Opus 5.5 (58.6). API pricing is set at $1.36 per million input tokens and $4.18 per million output tokens, undercutting Claude Opus 5.5 by a significant margin. Mistral has committed to releasing open weights by October 31, 2026.

Why it matters: 'Le Chonk' positions Mistral as a leading open-source AI lab in the US and Europe, though it lags behind Claude on automation and coding tasks. The model's open weights promise, competitive pricing, and strong performance on select benchmarks could shift the balance in enterprise AI adoption, especially for organizations seeking alternatives to closed US or Chinese models. However, the framing of 'best open weights' notably excludes top-performing Chinese models, highlighting a fragmented global AI landscape.

Source: Decrypt, VentureBeat