Mistral AI Pricing 2026: API Token Rates, Plans, and Full Model Comparison

If you are searching for Mistral AI pricing or Mistral API pricing in 2026, this guide covers every model, every rate, and every discount. Mistral offers some of the most competitive API pricing in the LLM market -- Mistral Small at $0.10 per million input tokens is over 25x cheaper than OpenAI's GPT-5.5 at $5.00. Whether you are evaluating Mistral for a production application or comparing it against OpenAI and Anthropic, the numbers below are current as of May 2026.

---

Mistral AI API Pricing at a Glance

Language Models -- Per Million Tokens

Specialised Models

---

Mistral Free Tier -- 1 Billion Tokens Per Month

Mistral's free API tier ("Experiment Plan") is one of the most generous in the industry.

What the free tier includes:

• 1 billion tokens per month across all models
• Access to all models including Mistral Large 2 and Codestral
• No credit card required -- phone number verification only
• Standard rate limits apply

At Mistral Small's production rate of $0.10/M input, 1 billion free tokens represents $100 in monthly value. For developers building and testing applications, this is a genuinely usable free tier -- not a trial.

Comparison to competitors:

• OpenAI: limited to ~$5 in new-user credits (or no free tier, reports vary)
• Anthropic: no free API tier -- subscription plans only
• Mistral: 1 billion tokens/month, no card required

---

Mistral Model Breakdown: Which Model to Use

Mistral Nemo -- $0.02/$0.04

Mistral Nemo is the cheapest capable model in Mistral's lineup. At $0.02 per million input tokens, running a million requests costs less than most other platforms charge per thousand.

Best for: High-volume classification, keyword extraction, simple summarisation, routing tasks, and any application where throughput and cost matter more than reasoning depth.

Context window: 128K tokens.

Ministral 3B -- $0.04/$0.04

Ministral 3B is a compact model designed for edge deployment and high-speed inference. Input and output are the same rate ($0.04/M), which simplifies cost modelling for output-heavy workloads.

Best for: Real-time autocomplete, low-latency chat, on-device inference scenarios, and applications needing the lowest possible latency.

Mistral Small 3.1 -- $0.10/$0.30

Mistral Small is the best-value model in Mistral's range for most production applications. At $0.10/M input and $0.30/M output with a 128K context window, it handles complex writing, summarisation, and structured extraction at a fraction of frontier model costs.

Best for: Customer support automation, document processing, structured data extraction, content generation, and any workload requiring real reasoning capability at low cost.

Cost comparison: Mistral Small ($0.10/$0.30) vs GPT-4o ($2.50/$10.00) -- Mistral Small is 25x cheaper on input and 33x cheaper on output for similar-tier general tasks.

Codestral -- $0.30/$0.90

Codestral is Mistral's specialised code model with a 256K context window -- the largest in Mistral's lineup. It is optimised specifically for code completion, debugging, and refactoring.

Best for: Code completion tools, IDE integrations, automated code review, test generation, and any developer-facing application where code is the primary modality.

Context advantage: The 256K context window lets you pass entire codebases or large files in a single prompt -- critical for code tasks that require understanding across multiple files.

Mistral Medium 3 -- $0.40/$2.00 and Mistral Medium 3.5 -- $1.50/$7.50

The Medium tiers sit between Small and Large on capability and cost. Medium 3 at $0.40/$2.00 is a cost-effective step up from Small when more reasoning depth is needed. Medium 3.5 at $1.50/$7.50 approaches Large capability at a lower price.

Best for: Complex analysis, multi-step reasoning, nuanced writing tasks, and applications where Small is insufficient but Large is cost-prohibitive.

Mistral Large 2 -- $2.00/$6.00

Mistral Large 2 is Mistral's flagship model. At $2.00/$6.00 per million tokens it is significantly cheaper than Claude Opus 4.7 ($5.00/$25.00) and GPT-5.5 ($5.00/$30.00) for frontier-tier reasoning tasks.

Best for: Complex multi-step reasoning, advanced code generation, nuanced long-form analysis, and applications where maximum Mistral capability is required.

Pixtral Large -- $2.00/$6.00

Pixtral Large is Mistral's multimodal model -- it processes both images and text at the same price as Mistral Large 2, with a 131K context window.

Best for: Document understanding with images, chart and diagram analysis, visual question answering, and any application combining text and image inputs.

---

Le Chat: Mistral Consumer Plans

Le Chat is Mistral's consumer AI assistant (the equivalent of ChatGPT or Claude.ai).

Le Chat Pro includes access to Mistral's best models, higher message limits, and web search integration. At $14.99/month it is cheaper than ChatGPT Plus ($20/month) and Claude Pro ($20/month).

---

Mistral API vs OpenAI vs Anthropic: Full Price Comparison

Budget / Fast Models

Mistral Nemo and Ministral 3B are the cheapest capable models in this comparison -- significantly undercutting even GPT-4o Mini. Mistral Small at $0.10/$0.30 is 1.5x cheaper on input than GPT-4o Mini and over 10x cheaper than Claude Haiku 4.5.

Mid-Tier Models

Mistral's mid-tier models are dramatically cheaper than GPT-4o and Claude Sonnet 4.6 for many use cases.

Frontier Models

Mistral Large 2 at $2.00/$6.00 is the cheapest frontier-class model in this comparison -- 2.5x cheaper on output than GPT-4o, and 5x cheaper on output than Claude Opus 4.7.

The trade-off: Mistral Large 2's benchmark performance on complex reasoning trails behind GPT-5.5 and Claude Opus 4.7 on some tasks. For applications where Mistral's capability meets the bar, the cost advantage is significant.

---

Mistral API Cost Examples

High-Volume Customer Support (10M requests/month, ~500 tokens each)

At scale, the cost gap between Mistral and OpenAI/Anthropic is enormous.

Code Completion Application (1M code completions/month, ~300 tokens each)

Codestral is purpose-built for code and significantly cheaper than using a general-purpose frontier model.

Document Analysis (100K documents/month, ~2,000 tokens each)

---

Who Should Use Mistral API

Choose Mistral if:

• Cost is the primary constraint and you need capable language models at the lowest price
• You are building a code-focused application (Codestral's 256K context and specialised training make it the best-value code model)
• You want a generous free tier to prototype and test without a credit card
• You need EU-based infrastructure (Mistral is a French company with European data residency)
• You want frontier-tier capability at lower cost than OpenAI or Anthropic (Mistral Large 2)

Choose OpenAI or Anthropic if:

• You need the absolute highest reasoning quality (GPT-5.5 and Claude Opus 4.7 outperform Mistral Large on complex tasks)
• You need embeddings (Mistral has no embedding model competitive with OpenAI's)
• You need image generation (Mistral has no image generation)
• You need the most mature developer ecosystem and tooling

---

Frequently Asked Questions

What is Mistral AI API pricing in 2026? Mistral Nemo costs $0.02/$0.04 per million tokens. Mistral Small 3.1 costs $0.10/$0.30. Codestral costs $0.30/$0.90. Mistral Large 2 costs $2.00/$6.00. All prices are input/output per million tokens.

Does Mistral have a free API tier? Yes -- the Experiment Plan gives 1 billion tokens per month across all models with no credit card required, just phone verification.

Is Mistral cheaper than OpenAI? Significantly so for most models. Mistral Small ($0.10/$0.30) is roughly 25x cheaper than GPT-5.5 ($5.00/$30.00) on input. Mistral Large 2 ($2.00/$6.00) is cheaper than GPT-4o ($2.50/$10.00) on both input and output.

What is Codestral used for? Codestral is Mistral's code-specialised model with a 256K context window. It is designed for code completion, debugging, refactoring, and test generation -- and is significantly cheaper than using general frontier models for code tasks.

How does Mistral compare to Claude API pricing? Mistral Small ($0.10/$0.30) is 10x cheaper than Claude Haiku 4.5 ($1.00/$5.00). Mistral Large 2 ($2.00/$6.00) is cheaper than Claude Sonnet 4.6 ($3.00/$15.00) on both input and output.

What is Le Chat Pro? Le Chat is Mistral's consumer AI assistant. Le Chat Pro costs $14.99/month -- cheaper than ChatGPT Plus ($20/month) and Claude Pro ($20/month) -- and includes access to Mistral's best models with higher message limits.

---

Building an AI startup? List it on Startup Launch Page and reach investors and early adopters actively looking for what you're building.