Mistral Large 4
Mistral's trillion-parameter flagship — Europe's strongest open-weight model, in public preview now.
About
Mistral Large 4 (ML4, nicknamed "le Chonk") is a natively multimodal MoE model with 1.05T total parameters and 52B active, announced October 6, 2026 as a public API preview. Mistral promises open weights by the end of October 2026 after red-teaming with security partners — as of Oct 13 the weights are not yet public and the license is unpublished.
The company emphasizes European AI sovereignty (trained and served from its own EU datacenters). Note: Mistral's own model card and announcement disagree on active parameter counts (49B vs 52B) — the card's 52B is used here.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Mistral-reported (all vendor claims; no independent testing published at preview):
- Cybench: 93%
- CyberGym-E2E: 82%
- Artificial Analysis Intelligence Index: 38 (claimed highest for a non-Chinese open-weight model)
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Mistral is a solid, established lab with clear preview pricing and an aggressive sale rate. But this entry is inherently provisional: open weights were promised, not delivered; the license is unpublished; all benchmarks are vendor-claimed with no independent runs; and Mistral's own model card and announcement disagree on active parameter counts.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 8.0/10
- Maintenance · recency, releases, issue load
- 9.0/10
- Openness · license + docs
- 6.5/10
- Manual review · sanity read + red-flag check
- 6.0/10
- · provider-tier estimate — thinner evidence (no public adoption metric)
- · released 0 days ago
- · manual review: provisional: no weights, no license, param-count discrepancy
- · capped at 7: key claims are provider-only, not yet independently verifiable