One-sentence summary
This dataset compiles comparable figures for Mistral Large 4 (Le Chonk), opened October 29, 2026, against GLM-5.3, Kimi K3 and DeepSeek V4.1 Flash on coding agents, terminal tasks, cybersecurity, intelligence index and API pricing, from Mistral's model card and third-party evals.
Data Notes
- Coding and cybersecurity benchmarks are as published in Mistral's official model card; Intelligence Index v5 follows Artificial Analysis.
- API prices are USD per million tokens; Mistral Large 4 endpoint is mistral-large-4.
- Reference models are only those cited in official benchmarks; unverified fields are marked unverified.
Core Data
| model | Total params | Active params | Context | Coding Agent Index | Terminal-Bench 4.0 | Cyber Index | Intelligence v5 | Input $/M | Output $/M |
|---|---|---|---|---|---|---|---|---|---|
| Mistral Large 4 | ~1.05T | ~52B/token | 1M | 48% | 41% | 51% | 26 | 0.70 | 2.10 |
| GLM-5.3 | Unverified | Unverified | Unverified | ~48% | Reference | Unverified | Higher than Mistral | Unverified | Unverified |
| Kimi K3 | Unverified | Unverified | Unverified | Reference | Reference | Unverified | Higher than Mistral | Unverified | Unverified |
| DeepSeek V4.1 Flash | Unverified | Unverified | Unverified | Reference | Reference | Unverified | Higher than Mistral | Unverified | Unverified |
Field Notes
- Coding Agent Index: Artificial Analysis coding-agent score; higher is better.
- Terminal-Bench 4.0: terminal task completion; Mistral improved 13 points from preview.
- Cyber Index: Artificial Analysis cybersecurity score for discovery and patching.
- Intelligence v5: Artificial Analysis Intelligence Index v5; Mistral's 26 is the top western open-weight score.
- Cached input price: $0.07 per million tokens.
Sources
- Unite.AI: Mistral Le Chonk
- IT Home: Mistral open-sources Large 4 weights