Model ComparisonMistral AI

Mixtral 8×22B

Mixtral 8×22B — Mistral's largest open-weight model with 141B total parameters.

VisionTool usePrompt cachingReasoningMedium
Pricing (verified 2026-08-19)
TypePrice / 1M tokensNotes
Input$1.20Per 1M input tokens
Output$1.20Per 1M output tokens
Source: Mistral AI official pricing page · Last verified 2026-08-19
Context Window
64K tokens
48K words · 0 pages of text

At $1.20/M input via Together AI, Mixtral 8×22B delivers performance comparable to GPT-4-class models at a fraction of the cost. The largest open-weight MoE model with strong multilingual capability.

Best for

  • Complex reasoning and code generation at scale
  • Multilingual applications (strong on French, Spanish, German, Italian, Portuguese)
  • Teams self-hosting on H100 / multi-A100 clusters
  • Function calling and structured output at lower cost than proprietary models

Not ideal for

  • Resource-constrained teams (requires 48+ GB VRAM at Q4 for local deployment)
  • Tasks requiring frontier-level reasoning (GPT-5.4 or o3 are stronger)

Real cost scenarios

ScenarioEst. monthly costBreakdown
Multilingual content — 1M queries/month at 2k tokens$3,6001.2B input at $1.20/M + 800M output at $1.20/M

Estimates with typical token distributions. Use the Prompt Cost Calculator for your exact workload.

How to cut costs on Mixtral 8×22B

  1. Self-hosting on a 4× A100 80GB cluster costs ~$7,300/month (RunPod) and gives unlimited inference — break-even vs API is roughly 3M tokens/hour utilization.

Calculate your Mixtral 8×22B costs

Prompt Cost Calculator
Exact cost for your prompt
Agent Cost Calculator
Multi-step agent loops
Prompt Caching Calculator
Caching savings
Conversation Cost
Chat history strategies

Other Mistral AI models

Mistral Large 2$2.00/MMistral Small 3$0.10/MMinistral 8B$0.10/M

Compare Mixtral 8×22B

deepseek v3 vs claude sonnet 5

Frequently Asked Questions

Is Mixtral 8×22B better than Llama 3 70B?

On most benchmarks, Mixtral 8×22B outperforms Llama 3.1 70B on reasoning and multilingual tasks. However, Llama 3.3 70B (the newer model) is competitive on English tasks at lower cost. For multilingual use, Mixtral 8×22B is still preferred.

Pricing verified 2026-08-19 from Mistral AI official pricing. All prices in USD per 1M tokens. See methodology and price changelog.