API Pricing
Mistral API Pricing 2026:
All Models, Costs & Comparison
Mistral AI offers some of the most affordable LLM APIs in 2026, with models ranging from ultra-cheap Mistral 7B to powerful Mistral Large. Complete pricing breakdown and real-world cost examples.
9 min read·Updated March 2026
Mistral Pricing Highlights
$0.10
Mistral 7B input/1M
$2.00
Mistral Large input/1M
$1.00
Mistral Small input/1M
32K
standard context window
Mistral API Models & Pricing 2026
| Model | Input (per 1M) | Output (per 1M) | Context | Best For |
|---|---|---|---|---|
| Mistral Large 2 | $2.00 | $6.00 | 128K | Complex reasoning, coding |
| Mistral Small 3 | $0.10 | $0.30 | 32K | High-volume tasks |
| Mistral Nemo | $0.15 | $0.15 | 128K | Best budget option, same in/out price |
| Codestral | $0.20 | $0.60 | 256K | Code generation, fill-in-middle |
| Mistral Embed | $0.10 | N/A | 8K | Embeddings, RAG |
| Pixtral Large | $2.00 | $6.00 | 128K | Vision + text tasks |
Mistral vs OpenAI vs Anthropic: Price Comparison
| Tier | Mistral | OpenAI | Claude |
|---|---|---|---|
| Budget | Nemo $0.15/M | 4o mini $0.15/M | Haiku $0.25/M |
| Balanced | Small 3 $0.10/M | GPT-4o $2.50/M | Sonnet $3.00/M |
| Premium | Large 2 $2.00/M | GPT-4o $2.50/M | Sonnet $3.00/M |
| Code-specific | Codestral $0.20/M | GPT-4o $2.50/M | Sonnet $3.00/M |
Key finding: Mistral Large is 4× cheaper than GPT-4o for input tokens at comparable quality levels. For most production workloads, Mistral represents significant cost savings.
Mistral Free Tier
Mistral offers free API access for experimentation via La Plateforme (console.mistral.ai). Free tier limits:
- Rate limits: ~1 request/second
- Free credits for new accounts
- Models available: Mistral Small, Nemo, Codestral
When to Choose Mistral Over OpenAI
- Cost-sensitive applications: Mistral Large at $2/M vs GPT-4o at $2.50/M for comparable quality
- European data residency: Mistral is French — EU-based infrastructure option
- Code generation: Codestral at $0.20/M with 256K context is exceptional value
- Open-source flexibility: Mistral 7B and Mixtral are open source — self-host for $0 variable cost
Real-World Mistral Cost Examples
Customer Support Bot (100K conversations/month)
- Each conversation: 600 tokens in + 400 tokens out = 1,000 tokens
- Total: 60M input + 40M output tokens/month
- Mistral Nemo: $9 + $6 = $15/month
- GPT-4o: $150 + $400 = $550/month
- Mistral saves $535/month (97% cheaper)
Code Review Tool (10K PRs/month)
- Average: 3,000 tokens in + 500 tokens out
- Codestral: $6 + $3 = $9/month
- GPT-4o: $75 + $50 = $125/month
- Codestral: 93% cheaper with 256K context
Self-Hosting Mistral Models
For very high volumes, self-hosting Mistral open-source models eliminates per-token costs entirely:
- Mistral 7B: ~4GB VRAM, fits on single GPU (RTX 4090 ~$1,600)
- Mixtral 8x7B: ~50GB VRAM, requires A100 80GB (~$2/hour on RunPod)
- Break-even vs API: typically at 50M+ tokens/month
Compare Mistral vs OpenAI Costs
Enter your monthly usage to see exactly how much you save with Mistral.
AI Cost Calculator