🇫🇷 France🧠 GeneralistReleased April 2024
Mixtral 8x22B
Mistral AI
Open-sourceFreeSelf-hostRGPD
Verdict
A solid pick for self-hosting and full data control.
Overview
Mixtral 8x22B is one of the first credible open-source MoE models, under Apache 2 license (unconstrained use). Older than the new lineup, but still relevant for absolute sovereignty needs (no license constraint).
Skill profile
Arena Elo1145
Verified benchmarks
MMLU77.5%
HumanEval76%
What do these scores mean?
MMLU77.5%very good
General knowledge across dozens of academic subjects.
HumanEval76%very good
Python code generation from specifications.
Strengths
- Strong French quality
- Has a free tier
- Open-source and self-hostable
- Competitive input pricing
Limitations
- Lagging on Arena
Who is it for
A good fit if…
- you want to control cost or self-host
- you work in French
- you have GDPR constraints
Skip it if…
Ideal use cases
- Fully free self-host
- European multilingual
- Academic research
- Sovereignty
Access & availability
Paid API per tokenFree tier availableSelf-hostable (open weights)
Key specifications
Context
64K
Input price
$0.60 $/M
Output price
$0.60 $/M
Speed
100 tok/s
Price not auditedBenchmarks not audited
Estimate your monthly cost
Per-token API pricing$22/ month with Mixtral 8x22B
For the same usage
- GPT-5$108+386%
- Claude Sonnet 4.6$196+781%
- Grok 4$256+1051%
Indicative estimate based on standard API rates (excluding caching, batch and volume discounts). Always check the official pricing before committing.
Privacy
GDPR-compliantHosting : EU/selfAnonymizable data
Advanced data · for expertsArchitecture, modalities, detailed cost, full benchmarks▾
Advanced data · for experts
Architecture, modalities, detailed cost, full benchmarks
Architecture
MoE
Size
141B (39B actifs MoE)
Cutoff date
Dec 2023
Inputs
text
Outputs
text, code
License
apache-2
Hosting
EU/self
Hallucination score
3/5
Estimated cost (API)
Typical exchange (~3k in / 1k out)
≈ $0.0024
1M in + 1M out
≈ $1.20
All benchmarks
| Arena Elo | 1145 |
| MMLU | 77.5% |
| GPQA | — |
| HumanEval | 76.0% |
| SWE-Bench | — |
| MATH | — |