Claude Fable 5$22.000/MClaude Opus 4.8$11.000/MClaude Opus 4.7$11.000/MClaude Opus 4.6$11.000/MClaude Opus 4.5$33.000/MClaude Sonnet 3.7$6.600/MClaude Opus 3$33.000/MClaude 2.1$12.800/MClaude 2$12.800/MGPT-5.5$12.500/MGPT-5.2$5.425/MGPT-5.2-Codex$5.425/MGPT-5$3.875/MGPT-4.5$97.500/MGPT-4 Turbo Preview$16.000/MGPT-4$39.000/MGPT-4-32k$78.000/Mo3$19.000/Mo3-mini$2.090/Mo4-mini$2.090/Mo1$28.500/Mo1-mini$5.700/Mo1-preview$28.500/MGemini 3.5 Pro$5.000/MGemini 3.1 Pro$5.000/MGemini 3 Pro$5.000/MGemini 2.5 Pro$3.875/MGemini 1.5 Pro$2.375/MGemini 1.0 Ultra$12.000/MGemini 1.0 Pro$0.800/MClaude Fable 5$22.000/MClaude Opus 4.8$11.000/MClaude Opus 4.7$11.000/MClaude Opus 4.6$11.000/MClaude Opus 4.5$33.000/MClaude Sonnet 3.7$6.600/MClaude Opus 3$33.000/MClaude 2.1$12.800/MClaude 2$12.800/MGPT-5.5$12.500/MGPT-5.2$5.425/MGPT-5.2-Codex$5.425/MGPT-5$3.875/MGPT-4.5$97.500/MGPT-4 Turbo Preview$16.000/MGPT-4$39.000/MGPT-4-32k$78.000/Mo3$19.000/Mo3-mini$2.090/Mo4-mini$2.090/Mo1$28.500/Mo1-mini$5.700/Mo1-preview$28.500/MGemini 3.5 Pro$5.000/MGemini 3.1 Pro$5.000/MGemini 3 Pro$5.000/MGemini 2.5 Pro$3.875/MGemini 1.5 Pro$2.375/MGemini 1.0 Ultra$12.000/MGemini 1.0 Pro$0.800/M
BETA
Subscribe
← Back to Index·All Mistral models
Mistral
MistralEfficientLIVE INDEX

Mixtral 8x22B Instruct

text+file->text

Largest open-weights Mixtral MoE. Cheap-to-serve frontier-ish quality before Llama 3.1 405B and DeepSeek V3 took the open-weights lead.

Mixtral 8x22B Instruct is a efficient AI model from Mistral. It costs $2.000 per million input tokens and $6.000 per million output tokens (blended $3.200/M), with a 66K-token context window.

Profile inherited from upstream Mixtral 8x22B ↗ — this is a hosted variant of the same open-weights model.

Released Apr 2024Modalities textOfficial model page ↗Provider pricing ↗API docs ↗Pricing source ↗Compare with another model →Estimate monthly cost →
INPUT
$2.000/M
per million input tokens
OUTPUT
$6.000/M
per million output tokens
BLENDED 70/30
$3.200/M
default reference rate · how it's calculated →
CONTEXT
66K
65,536 tokens
What it's good at
  • Open-weights MoE
  • Cheap inference per active parameter
  • 64K context
Typical use cases
  • Self-hosted chat at scale
  • Fine-tune base
Benchmarks
vs. best public score
Scores inherited from Mixtral 8x22B — this is a hosted variant of the same open-weights model, so the underlying benchmark scores are identical.
MMLU78%
Multitask academic knowledge across 57 subjects.
GPQA Diamond36%
Graduate-level science questions, "Google-proof".
MATH42%
High-school competition math problems.
HumanEval76%
Python function synthesis from docstrings.
LMArena Elo1198 Elo
Crowd-sourced head-to-head preference Elo rating.
Hand-curated from each provider's published reports and public leaderboards. Methodology varies across sources — treat as directional rather than authoritative.
More from Mistral
See all 15 →
Mistral Large 3
Frontier · 256K ctx
in $0.500/Mout $1.500/M
Mistral Medium 3.5
Balanced · 256K ctx
in $1.500/Mout $7.500/M
Mistral Small 4
Efficient · 128K ctx
in $0.100/Mout $0.300/M
Mistral Large 2
Frontier · 128K ctx
in $2.000/Mout $6.000/M
Pixtral Large
Vision · 128K ctx
in $2.000/Mout $6.000/M
Pixtral 12B
Compact Vision · 128K ctx
in $0.150/Mout $0.150/M
Frequently asked questions

How much does Mixtral 8x22B Instruct cost?

Mixtral 8x22B Instruct costs $2.000 per million input tokens and $6.000 per million output tokens, for a blended reference rate of $3.200 per million tokens.

What is Mixtral 8x22B Instruct's context window?

Mixtral 8x22B Instruct supports up to 66K tokens of context (65,536 tokens).

What is Mixtral 8x22B Instruct best for?

Mixtral 8x22B Instruct is well suited to Open-weights MoE, Cheap inference per active parameter and 64K context.

Who makes Mixtral 8x22B Instruct?

Mixtral 8x22B Instruct is developed and served by Mistral. It was released in Apr 2024.