Mixtral 8x22B
Mistral's 141B MoE under Apache-2.0 — rare permissiveness at this scale.
About
Mixtral-8x22B-Instruct is Mistral's large mixture-of-experts model: 8 experts of 22B each (141B total, 39B active per token via top-2 routing), 64K context, strong multilingual and reasoning performance for its 2024 generation. The notable fact: it's Apache-2.0 licensed — genuinely open source at 141B parameters, a rarity at this scale. 41.2k HF downloads, not gated. The catch is weight: 281GB of weights needs serious GPU infrastructure (or a hosted endpoint). Superseded by Mistral Large 2/3/4, but the permissive license keeps it relevant for commercial self-hosting.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported (Mistral, Apr 2024): strong vs GPT-3.5-class on MMLU, HumanEval, MATH; efficient MoE inference — provider-reported. Historical; treat 2024 comparisons as dated.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (verified tier, metrics-v1). Evidence: HF downloads 41,204 (peer-normalized); released 910 days ago. Manual review: Mistral's 141B MoE (8x22B, 2 active) — genuinely Apache-2.0 at this scale; 281GB weights = serious GPU needs; superseded by Mistral Large gens. License: Apache-2.0.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 4.0/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · HF downloads 41,204 (peer-normalized)
- · released 910 days ago
- · manual review: Mistral's 141B MoE (8x22B, 2 active) — genuinely Apache-2.0 at this scale; 281GB weights = serious GPU needs; superseded by Mistral Large gens