Mistral AI sells its models pay-as-you-go through its API, and several of them are also published as open weights under Apache 2.0. The lineup is changing: Mistral Large 3 and Medium 3.5 were deprecated on Oct 31, 2026, and Mistral points users to Mistral Large 4 instead.
Mistral Small 4 costs $0.15 input / $0.60 output per 1M tokens. Ministral 3 14B charges $0.20 for both input and output, Codestral is $0.30 / $0.90, and Mistral Large 3 is $0.50 / $1.50. Mistral says Batch processing cuts the price by 50%.
| Model | Tier | Input / 1M | Output / 1M | Cache hit / 1M | Context | Notes |
|---|---|---|---|---|---|---|
| Mistral Large 4 | Mid-tier | $0.70 | $2.10 | $0.07 | 1M | New Mistral flagship (Oct 6, 2026; public preview); open-weight 1.05T MoE (52B active), multimodal; replaces Large 3 and Medium 3.5 |
| Mistral Large 3 | Mid-tier | $0.50 | $1.50 | — | 256K | Former Mistral flagship; deprecated Oct 31, 2026, retiring Apr 30, 2027 — replaced by Mistral Large 4 |
| Mistral Small 4 | Budget | $0.15 | $0.60 | $0.015 | 256K | Mid-tier general; replaced Small 3.1 |
| Codestral | Specialist | $0.30 | $0.90 | $0.03 | 128K | Code generation |
| Ministral 3 14B | Budget | $0.20 | $0.20 | $0.02 | 256K | Budget multimodal; replaced Pixtral 12B |
| Mistral Medium 3.5 | Mid-tier | $1.50 | $7.50 | — | 256K | Frontier-class multimodal; agentic & coding focus; deprecated Oct 31, 2026, retiring Apr 30, 2027 — replaced by Mistral Large 4 |
| Model | Support chatbot 10,000 conversations × (2,000 input + 400 output tokens) | Document Q&A 1,000 questions × (30,000 input + 800 output tokens) | Coding agent session 4M input tokens (90% cache hits) + 150,000 output tokens |
|---|---|---|---|
| Ministral 3 14B | $4.80 | $6.16 | $0.18 |
| Mistral Small 4 | $5.40 | $4.98 | $0.20 |
| Codestral | $9.60 | $9.72 | $0.36 |
| Mistral Large 3 | $16.00 | $16.20 | $2.23 |
| Mistral Large 4 | $22.40 | $22.68 | $0.85 |
| Mistral Medium 3.5 | $60.00 | $51.00 | $7.13 |
Standard rates per 1M tokens from data on this site. The coding-agent row assumes 90% of input is read from the prompt cache (full input price where a model has no cache rate). Cache-write fees, batch discounts and long-context surcharges listed in the notes are not included. Updated November 2, 2026.
| Model | Provider | Tier | Input / 1M | Output / 1M | Context |
|---|---|---|---|---|---|
| Mistral Large 3 | Mistral | Mid-tier | $0.50 | $1.50 | 256K |
| DeepSeek V4-Pro | DeepSeek | Mid-tier | $0.66 | $1.98 | 1M |
| Grok 4.3 | SpaceXAI | Mid-tier | $1.25 | $2.50 | 1M |
| Gemini 3.8 Flash | Mid-tier | $0.75 | $3.75 | 1M | |
| Gemini 3.6 Flash | Mid-tier | $0.75 | $3.75 | 1M | |
| Ministral 3 14B | Mistral | Budget | $0.20 | $0.20 | 256K |
| Gemini 2.5 Flash-Lite | Budget | $0.10 | $0.40 | 1M | |
| GPT-6 Luna | OpenAI | Budget | $0.10 | $0.50 | 1M |
| Claude Haiku 5.5 | Anthropic | Budget | $0.10 | $0.50 | 1M |
| DeepSeek V4.1-Flash | DeepSeek | Budget | $0.15 | $0.60 | 1M |
Mistral's cheapest model per tier (bold) against the four cheapest same-tier models from other providers, by output price.
Mistral's model docs list Mistral Large 3 (mistral-large-2512) and Mistral Medium 3.5 (mistral-medium-3-5) as deprecated on Oct 31, 2026, with retirement on Apr 30, 2027, and name Mistral Large 4 as the replacement. Large 4 isn't in our rate tables yet. Mistral's pricing page lists it above Large 3 but well below Medium 3.5 ($1.50 / $7.50) per token, so Medium users should see their bill fall after the move, while Large 3 users should budget for a modest rise.
Mistral Large 3's $1.50 output is a fraction of other mid-tier models: Claude Sonnet 5.5 and GPT-6.1 Sol charge $10.00 and Gemini 3.8 Flash $3.75, and its replacement, Large 4, is also listed well under those rates. Ministral 3 14B prices input and output the same, which suits generation-heavy jobs. For input-heavy work like document Q&A, Small 4 is cheaper ($0.15 vs $0.20 input). Pick between the two by your input-to-output ratio, not by model size.
Our dataset has no cache-hit rates for Mistral, so the coding-agent cost above bills every input token at full price. That makes Mistral look worse than it is next to providers with cheap cache hits. Mistral's pricing page says cached input tokens cut input cost by up to 90% on repeated prompts, and Batch processing cuts the price by 50%. For a cached, asynchronous workload, the real bill can be well under the table figure.
Mistral's docs list Mistral Small 4 and Ministral 3 14B under the Apache 2.0 license, so you can prototype on the API and later self-host the same model if volume or data-residency rules justify it. Codestral is a commercial Premier model with no open-weights licence listed, so plan to stay on the API for it.
Mistral's pricing page lists a Free plan that includes access to Mistral Studio and limited Vibe usage, but it doesn't state a free API token allowance. The paid Pro plan includes a monthly API credit. Plan on paying the per-token rates shown here for production traffic.
Codestral, Mistral's code-generation model, costs $0.30 per 1M input tokens and $0.90 per 1M output tokens with a 128K context window. That's cheaper per output token than Mistral Large 3. For code work across large repositories, also test Mistral Small 4, which Mistral describes as unifying instruct, reasoning and coding: it costs less ($0.15 / $0.60) and has a bigger 256K window.
Mistral Small 4 ($0.15 / $0.60) and DeepSeek V4.1-Flash ($0.15 / $0.60) list the same rates, but DeepSeek's are off-peak prices that double during weekday peak hours. DeepSeek's cache hits ($0.003) are also far cheaper than Mistral's, which its pricing page puts at a tenth of the input price. Mistral's pricing page doesn't describe time-of-day pricing. For steady traffic around the clock, compare your peak share before choosing.
How we choose: prices come from each vendor's official pricing page and are re-checked every week. We take no affiliate fees, so rankings aren't for sale.