Provider pricing hub
Mistral API Pricing (July 2026)
The Mistral API costs $1.50/$7.50 per million input/output tokens for the flagship Mistral Medium 3.5, $0.50/$1.50 for Mistral Large 3, and $0.15/$0.60 for Mistral Small 4, with the Ministral edge tiers from $0.10. Medium 3.5 ships open weights under a modified MIT license.
Mistral's price sheet is the flattest in the frontier-adjacent market: symmetric input/output rates on the Ministral tiers, sub-$2 output on everything except the Medium 3.5 flagship, and open weights across much of the lineup. The table below lists every Mistral model in BenchLM's registry, synced from Mistral's official model cards in Mistral AI Studio.
Mistral is also the strongest European lab — relevant if data-sovereignty rules shape your provider list. See the Mistral model rankings and the European AI landscape guide for the benchmark side of the decision.
Every Mistral API price per 1M tokens
| Model | Input $/M | Cached input $/M | Output $/M | Context |
|---|---|---|---|---|
| Mistral Medium 3.5 128B | $1.5 | $0.15 | $7.5 | 256K |
| Mistral Large 3 | $0.5 | $0.05 | $1.5 | 256K |
| Mistral Medium 3 | $0.4 | $0.04 | $2 | 128K |
| Ministral 3 14B | $0.2 | $0.02 | $0.2 | 256K |
| Mistral Small 4 | $0.15 | $0.015 | $0.6 | 256K |
| Ministral 3 8B | $0.15 | $0.015 | $0.15 | 256K |
| Ministral 3 3B | $0.1 | $0.01 | $0.1 | 256K |
Rates are Mistral's official model-card prices in Mistral AI Studio. Reasoning siblings share their family's published price and are collapsed into one row. Mistral does not publish a flat batch-discount tier, so batch columns are omitted.
Estimate your monthly Mistral API bill
| Model | Est. monthly cost |
|---|---|
| Mistral Medium 3.5 128B | $30 |
| Mistral Large 3 | $8 |
| Mistral Medium 3 | $8 |
| Ministral 3 14B | $2.4 |
| Mistral Small 4 | $2.7 |
| Ministral 3 8B | $1.8 |
| Ministral 3 3B | $1.2 |
Estimates use Mistral's standard per-token rates from the table above. For task-level presets and cross-provider comparison, use the full AI cost calculator.
How the Mistral lineup is priced
Mistral Medium 3.5 ($1.50/$7.50) is the April 2026 flagship — a 128B dense model that Mistral ships as open weights under a modified MIT license, so the API price buys convenience rather than exclusivity. Mistral Large 3 ($0.50/$1.50) is the value-priced workhorse, and Mistral Small 4 ($0.15/$0.60) is a single hybrid model covering instruct, reasoning, and coding at budget rates.
The Ministral 3 family (3B/8B/14B, $0.10 to $0.20 with symmetric input/output pricing) targets edge and embedded deployments — the symmetric rates make output-heavy workloads unusually cheap.
Mistral vs OpenAI and DeepSeek pricing
Against OpenAI, Mistral competes on output price: Large 3's $1.50 output is a quarter of GPT-5.6 Luna's $6 and a twentieth of Sol's $30, and even the Medium 3.5 flagship's $7.50 output undercuts every current OpenAI flagship. Against DeepSeek, the comparison is tighter — V4 Flash ($0.14/$0.28) beats Small 4 on output — but Mistral counters with EU hosting and the Ministral edge tiers.
Quality-per-dollar is the real question: check the price-vs-performance view and the compare hub before defaulting to a bigger name.
Which Mistral should you use?
Small 4 is the default starting point — one hybrid model for instruct, reasoning, and coding at $0.15/$0.60. Step up to Large 3 when quality gaps appear, and to Medium 3.5 for the hardest work or when you want the open-weights escape hatch (start on the API, move to self-hosting at volume — see the local LLM guide). Use Ministral 3 tiers for edge, embedded, and offline deployments where per-device inference matters more than peak capability.
Mistral API pricing FAQ
How much does the Mistral API cost?
Mistral Medium 3.5 costs $1.50/$7.50 per million input/output tokens, Mistral Large 3 $0.50/$1.50, Mistral Small 4 $0.15/$0.60, and the Ministral 3 edge family $0.10–$0.20 with symmetric input/output rates. The full table above is synced from Mistral's official model cards.
Is the Mistral API free?
The hosted API is pay-per-token, but much of the lineup — including the Medium 3.5 flagship — ships open weights, so self-hosting is a real zero-per-token alternative. See the open-source rankings for which Mistral rows are worth running yourself.
What is Mistral's cheapest model?
Ministral 3 3B at $0.10/$0.10 per million tokens is the cheapest hosted Mistral model, and its symmetric pricing makes it especially cheap for output-heavy tasks. Among the full-size models, Mistral Small 4 at $0.15/$0.60 is the budget pick.
How does Mistral pricing compare to OpenAI?
Mistral is consistently cheaper on output: Large 3's $1.50/M output is 4x below GPT-5.6 Luna and 20x below GPT-5.6 Sol, and Medium 3.5's $7.50 undercuts every current OpenAI flagship. OpenAI's top models score higher on BenchLM's leaderboard, so the choice is quality-per-dollar — see the OpenAI pricing hub.