Compare AI models

Specifications
Parameters
1.05T2.8T
Context window
1M1M
API pricing
Input price
$0.70$3.00
Output price
$2.10$15.00
Cached input price
$0.07$0.30
Cheapest input
$0.70Mistral$0.153Wafer
Cheapest output
$2.10Mistral$9.50Chutes
$245Mistral list price$535Wafer · mxfp4
Benchmarks
BullshitBench v2
80%74%
DeepSWE 1.1
61.7%69%
Terminal-Bench 4.0
41%23.03%
BrowseComp
84.5%91.2%
threejseval
10081517
Overview
CompanyMistralMoonshot AI
Release dateOct 6 2026Jul 16 2026
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

Kimi K3 leads Mistral Large 4 on 3 of the 5 benchmarks they both report. Mistral Large 4 is cheaper on both input and output: $0.70 vs $3.00 per million input tokens, and $2.10 vs $15.00 per million output tokens. Kimi K3 shipped 82 days before Mistral Large 4, so benchmark comparisons should account for the intervening progress.

Mistral Large 4 has 1.05T parameters, while Kimi K3 has 2.8T. Context windows are 1M (Mistral Large 4) vs 1M (Kimi K3).

On BullshitBench v2, Mistral Large 4 leads at 80% vs Kimi K3 at 74%. On DeepSWE 1.1, Kimi K3 leads at 69% vs Mistral Large 4 at 61.7%. On Terminal-Bench 4.0, Mistral Large 4 leads at 41% vs Kimi K3 at 23.03%. On BrowseComp, Kimi K3 leads at 91.2% vs Mistral Large 4 at 84.5%. On threejseval, Kimi K3 leads at 1517 vs Mistral Large 4 at 1008.