Mistral Large 4 on BenchGecko · open weights and a price cut of about half
Mistral released the weights of Mistral Large 4 on 2026-10-29 and cut its API price from $1.36 to $0.70 input and from $4.18 to $2.10 output per 1M tokens.
What shipped
Mistral AI released the weights of Mistral Large 4 on 2026-10-29, three weeks after the model entered public preview on 2026-10-06. The release is an improved checkpoint, trained with more reinforcement learning since the preview. Weights are on Hugging Face as mistralai/Mistral-Large-4-1T-A52B.
BenchGecko updated the model page today: open weights, new price, Hugging Face link.
Price
The Mistral API price changed on the same day:
- Input · $0.70 per 1M tokens, down from $1.36 (48.5% lower)
- Output · $2.10 per 1M tokens, down from $4.18 (49.8% lower)
- Cached input · $0.07 per 1M tokens, down from $0.14
For comparison, Mistral Large 3 2512 lists at $0.50 input and $1.50 output on BenchGecko. Price history will show the change after the next daily snapshot.
Model facts
- Parameters · about 1T total, 52B active per token
- Experts · 128 routed, 4 active per token, plus 1 shared
- Context window · 1,048,576 tokens
- Input · text and image
- Reasoning modes · low, medium, high (default)
- License · Mistral Ecosystem Open-Weight License, a modified MIT license. Companies that sell the model as a hosted service need a separate commercial license.
Scores reported by Mistral
These are self-reported by Mistral and recorded as vendor claims, not BenchGecko scores.
- Artificial Analysis Coding Agent Index · 48%
- Terminal-Bench 4.0 · 41%, up 13 points from the preview
- FrontierCode 2.0 (Cognition) · 19.7%
- Artificial Analysis Cyber Index · 51%
- BrowseComp · 84.5%
- FinWorkBench (Finch) · 73.3%
- B3 agent security, attacks resisted · 97.1%, up from 93.3% in preview
On BenchGecko
Mistral Large 4 has an average score of 40.6 across 21 benchmarks on BenchGecko. Those results were added during the preview, so they may not reflect the new checkpoint. BenchGecko will add independent results for the released weights as they appear.
Sources
Source: BenchGecko
Written automatically from BenchGecko's tracked data for this period; every number comes from the changelog, model, mindshare and pricing datasets. Checked against the source data before publishing.