AI Tools

What Does Mistral API Cost in 2026? Full Rate Card

Published 2026-09-08 · Researched and price-checked on official sources

Mistral API pricing in 2026 starts at $0.50 per million input tokens on Mistral Large 3, its open-weight flagship, with output at $1.50 per million. The company announced a EUR 3 billion Series D this week, and every rate below was verified on the official pricing page on September 8, 2026.

Key facts

What matters

Mistral Large 3 lists at $0.50 per million input and $1.50 per million output tokens on the standard tier, per the official pricing page.
Mistral Medium 3.5, the agentic flagship, costs $1.50 per million input and $7.50 per million output, three times Large 3's input rate despite the smaller name.
Batch processing cuts every Mistral rate by 50%, and cached input tokens cost up to 90% less than fresh input on repeated prompts.
Enterprise APIs on select models list at 75% above standard pricing, with regional data controls and SLAs, per Mistral's pricing page.
Mistral Small 4, labeled Apache 2.0 on the pricing page, lists at $0.15 per million input and $0.60 output for cost-sensitive projects.

What the EUR 3B raise changes for Mistral API buyers

Mistral announced a EUR 3 billion Series D at a post-money valuation above EUR 21 billion, the largest equity round ever completed by a European technology company, according to the official announcement. Samsung Electronics led the round, with Scaleup Europe Fund and PSG Equity as co-leads. The company frames the raise around its sovereign AI stack: open-weight models, infrastructure, and the products that bring them into production.

For a developer choosing an LLM API, the raise matters because it funds the frontier research and compute that keep Mistral's models current. It does not change the rate card. The prices below were captured from mistral.ai/pricing and mistral.ai/pricing/api on September 8, 2026.

Every Mistral text model rate, verified September 8, 2026

Mistral prices most models per million tokens, with input and output billed separately. The standard tier is the default; batch and priority tiers are separate. Model IDs follow the 'latest' alias convention shown on the pricing page, so the rates apply to the current version of each model family.

Mistral Medium 3.5: $1.50 input, $7.50 output per million tokens. Positioned for long-horizon tasks, synchronous tool-calling, and agentic coding.
Mistral Large 3: $0.50 input, $1.50 output. Open-weight, general-purpose flagship, multimodal and multilingual.
Mistral Small 4: $0.15 input, $0.60 output. Multimodal and multilingual, labeled Apache 2.0 on the pricing page.
Codestral: $0.30 input, $0.90 output. Premier coding model for low-latency completion and fill-in-the-middle tasks.
Ministral 3 (3B): $0.10 in and $0.10 out. Ministral 3 (8B): $0.15 in and $0.15 out. Ministral 3 (14B): $0.20 in and $0.20 out.
GLM 5.2 (third-party via Z.ai): $1.40 input, $4.40 output, with $0.14 cached input, listed as new on Mistral's API.

Mistral's hidden pricing structure: Medium 3.5 costs more than Large 3

The naming no longer maps to price. Mistral Medium 3.5 lists at $1.50 input and $7.50 output, while Mistral Large 3, the open-weight flagship, lists at $0.50 and $1.50. A developer who assumes 'Large' is the premium tier will budget wrong.

Mistral's own guidance on the pricing page says to reach for Medium for most tasks and coding, Small for cost-sensitive projects, OCR for documents, and Voxtral for audio. Medium 3.5 is the current agentic workhorse; Large 3 is the cheaper open-weight generalist. Check the model card before assuming the price follows the name.

Discount mechanics: batch, cached input, and priority tiers

Three levers change the effective rate on Mistral's API:

Batch: Mistral's batch API cuts the standard price by 50% for asynchronous, high-volume workloads.
Cached input: repeated prompt prefixes cost up to 90% less than fresh input, per the pricing FAQ.
Priority: a priority tier routes eligible requests through dedicated queues for faster, more predictable time to first token.

Enterprise APIs add regional data-processing controls and system-level SLAs at 75% above list pricing on select models, per the API pricing page. OCR is billed per 1,000 pages, speech models per minute, and tool APIs per call rather than per token.

How Mistral API pricing compares to GPT-6 Astra

The comparison anchor is OpenAI's current flagship. GPT-6 Astra lists at $10.00 per million input tokens and $50.00 per million output on the standard tier of OpenAI's pricing page, with $1.00 cached input. Mistral Large 3 at $0.50 input and $1.50 output is 20 times cheaper on input and more than 30 times cheaper on output, and Medium 3.5 still undercuts Astra by roughly 6.7x on both axes.

The tradeoff is ecosystem and capability, not headline price. OpenAI's models carry a wider tool and multimodal surface and its own cache-write economics, and GPT-6 Astra context and output limits exceed what Mistral publishes on its pricing page. Mistral's pitch is open weights: Large 3 and Small 4 carry OPEN labels, and Small 4 is explicitly Apache 2.0, which matters if you want to leave the API and self-host.

Which Mistral model should you pay for?

Match the model to the workload, not the ranking:

Agentic coding and long-horizon tasks: Medium 3.5 at $1.50/$7.50, per Mistral's own recommendation for most tasks and coding.
Open-weight generalist on a budget: Large 3 at $0.50/$1.50, or Small 4 at $0.15/$0.60 when you only need multimodal text.
High-frequency code completion: Codestral at $0.30/$0.90, built for low-latency fill-in-the-middle.
Edge and mobile workloads: Ministral 3 sizes from $0.10 in and out at 3B up to $0.20 at 14B.

All rates are USD list prices on the standard tier as of September 8, 2026, and are subject to change on the official page.

At a glance

Model	Input per 1M	Output per 1M	Positioning
Mistral Medium 3.5	$1.50	$7.50	Agentic, long-horizon tasks
Mistral Large 3	$0.50	$1.50	Open-weight generalist flagship
Mistral Small 4	$0.15	$0.60	Apache 2.0, cost-sensitive
Codestral	$0.30	$0.90	Code completion, FIM
Ministral 3 3B / 8B / 14B	$0.10 / $0.15 / $0.20	$0.10 / $0.15 / $0.20	Edge and mobile
GLM 5.2 (third-party)	$1.40	$4.40	Long-context agentic, $0.14 cached input

Mistral API standard-tier token rates, verified on mistral.ai/pricing/api, September 8, 2026

FAQ

Is there a free Mistral API tier?

Mistral's free Vibe plan includes $10 per month in API credits for Studio, and free plans exist for Vibe access. Paid consumer plans from Pro at $14.99 per month include more API credits and higher usage ceilings.

Which Mistral model is cheapest?

Ministral 3 at 3B is the cheapest at $0.10 per million tokens in and out. Among full-size models, Mistral Small 4 at $0.15 input and $0.60 output is the low-cost option, per the official pricing page.

Does batch processing really halve Mistral API costs?

Yes. Mistral's pricing FAQ states that batch processing reduces the price by 50% for high-volume asynchronous work, and cached input tokens reduce input cost by up to 90% on repeated prompts.