Mistral AI · Available now

Mistral Large 4 price per reply (Mistral AI, 2026): cost, context and open weights

Mistral Large 4 is Mistral AI’s open-weight multimodal model with about 1 trillion parameters and a 1M-token context window, first released as a public preview on October 6, 2026, priced at $0.70 per million input and $2.10 per million output tokens on Mistral’s API and OpenRouter.

Available nowLast updated:

Use it now

Chat

≈ 1 credit per reply · See all models

New accounts get 5 free credits once, valid for 7 days. They work on the lower-priced models here; the priciest ones need credits you bought.

Start chatting — 5 free credits
Maker
Mistral AI
Type
Text
Released
2026-10-06
OpenRouter id
mistralai/mistral-large-4-0

Per million tokens

Price

Mistral Large 4 · checked 2026-11-17
-InputOutputSource
Mistral AI — Standard$0.70$2.1docs.mistral.ai
Via OpenRouter (mistralai/mistral-large-4-0)$0.70$2.1openrouter.ai

Checked

Facts, with sources

  • Mistral’s launch post, dated October 6, 2026, announces a public preview of Mistral Large 4 on the preview API in Mistral Studio, with open weights to follow "by the end of the month".Source: mistral.ai · checked 2026-11-17
  • Mistral calls it "our largest and most capable model to date" and "natively multimodal"; its product card says "Our most capable model for agentic coding and cybersecurity." Mistral describes it as open-weight.Source: mistral.ai · checked 2026-11-17
  • A later Mistral post says "Today, we’re releasing the weights of this improved Mistral Large 4", and describes the October 6 version as the public preview. The post carries no date. Mistral says the new version had about three more weeks of reinforcement learning, and it now offers reasoning modes Low, Medium and High (default); Mistral’s own Terminal-Bench 4.0 figure rose 13 percentage points to 41% (the maker’s own figures, not tested by us).Source: mistral.ai · checked 2026-11-17
  • Mistral’s docs list Mistral Large 4 as a granular mixture-of-experts model with 1.05T total and 52B active parameters, a 1M-token context window and a 1.6B vision encoder, with function calling, structured outputs, batching and built-in tools. The launch post rounds the size to "1 trillion" total and 52B active. Mistral’s pages do not state a maximum output length.Source: docs.mistral.ai · checked 2026-11-17
  • Hugging Face model card (mistralai/Mistral-Large-4-1T-A52B): about 1T total parameters, 1M-token context, text and image input and text output, and a Pixtral-style vision encoder. The card gives 49B active parameters in its intro and "~52B" in its key features, so the active count differs inside the card.Source: huggingface.co · checked 2026-11-17
  • OpenRouter lists mistralai/mistral-large-4-0 with a 1,048,576-token context window, a maximum completion of 262,144 tokens on its top provider, and text and image input with text output (read November 17, 2026).Source: openrouter.ai · checked 2026-11-17 · second-hand
  • Mistral’s model page lists Mistral Large 4 at $0.7 per million input tokens, $0.07 cached input and $2.1 per million output tokens; the launch post and the open-weights post give the same figures ($0.70 / $2.10). Mistral’s changelog said launch pricing was 50% off for two weeks from the October 6 preview; we did not find a dated end of that offer.Source: docs.mistral.ai · checked 2026-11-17
  • OpenRouter’s listing for mistralai/mistral-large-4-0 shows $0.70 per million input tokens, $2.10 per million output tokens and $0.07 cached input, the same as Mistral’s own rate (read November 17, 2026).Source: openrouter.ai · checked 2026-11-17 · second-hand
  • Mistral says Mistral Large 4 is available in Mistral Studio, the Mistral API and OpenRouter, as weights on Hugging Face (deployable with vLLM or SGLang), on NVIDIA-hosted endpoints and an NVIDIA NIM container, in Microsoft Foundry, and in Vibe Code (chat.mistral.ai), with Mistral Forge for customisation. The API name is mistral-large-4 (mistral-large-4-0 also appears in its changelog); reasoning_effort accepts low, medium or high, defaults to high, and "none" now returns an error.Source: mistral.ai · checked 2026-11-17
  • The weights are under the Mistral Ecosystem Open-Weight License, which the model card calls "a modified version of the MIT license". It covers commercial and non-commercial use; a business running a "Model-as-a-Service" offering needs a separate commercial licence from Mistral. This is a summary, not legal advice.Source: huggingface.co · checked 2026-11-17

Nearest available

Models you can use now instead

Kimi K3

Moonshot AI

$0.19 / $16 (per million tokens)

  • Moonshot’s Kimi Code changelog lists Kimi K3 as released on July 16, 2026, as open source. Moonshot’s blog said the full weights would follow by July 27, 2026.Source: kimi.com · checked 2026-11-16
  • Moonshot says K3 is designed for long-horizon coding, knowledge work and reasoning, and that it is available on Kimi.com, Kimi Work, Kimi Code and the Kimi API.Source: kimi.com · checked 2026-11-16
  • Mixture-of-experts model with 2.8T total parameters and 104B active per token, a 1,048,576-token context window, and image and video understanding. In the API, thinking is always on, with reasoning effort low, high or max (default max), and maximum output tokens default to 131,072 and can be raised to 1,048,576.Source: platform.kimi.ai · checked 2026-11-16

$0.08 / $3.6 (per million tokens)

  • A secondary report (DataNorth) says Z.ai released GLM-5.3 on August 14, 2026; Z.ai’s docs page calls it "Z.ai’s latest flagship model" but states no date.Source: datanorth.ai · checked 2026-11-15 · second-hand
  • Z.ai’s docs say GLM-5.3 advances complex software engineering and agent capabilities, with an emphasis on coding and cybersecurity performance.Source: docs.z.ai · checked 2026-11-15
  • Context window 1M tokens, maximum output 128K tokens; text-only input.Source: docs.z.ai · checked 2026-11-15

FAQ

Questions

What is Mistral Large 4?

Mistral calls it "our largest and most capable model to date" and "natively multimodal", and its product card calls it the most capable model for agentic coding and cybersecurity. Mistral’s docs list a granular mixture-of-experts design with 1.05T total and 52B active parameters. You can chat with it on this page on credits.

How much does Mistral Large 4 cost?

Mistral’s model page lists $0.7 per million input tokens, $0.07 cached input and $2.1 per million output tokens; OpenRouter shows $0.70 / $2.10 as well (read November 17, 2026). Mistral’s changelog said launch pricing was 50% off for two weeks from October 6, and we did not find a dated end of that offer. On this site a typical reply to this model costs about 1 credit (a credit is about $0.01 in the $5 pack), billed by its real token cost, at least 1 credit.

How do I use Mistral Large 4?

Through Mistral Studio and the Mistral API (model name mistral-large-4), through OpenRouter as mistralai/mistral-large-4-0, in Vibe Code at chat.mistral.ai, on NVIDIA-hosted endpoints, in Microsoft Foundry, or by running the Hugging Face weights with vLLM or SGLang. You can also chat with it in the box on this page.

What is the context window of Mistral Large 4?

Mistral’s docs and the Hugging Face model card give a 1M-token context window; OpenRouter lists 1,048,576 tokens and a maximum completion of 262,144 tokens on its top provider. Mistral’s own pages do not state a maximum output length.

Is Mistral Large 4 open source?

Its weights are open on Hugging Face (mistralai/Mistral-Large-4-1T-A52B) under the Mistral Ecosystem Open-Weight License, which the model card calls a modified version of the MIT license. It allows commercial and non-commercial use, but a business running a "Model-as-a-Service" offering needs a separate commercial licence from Mistral. Read the licence before commercial use.

Is Mistral Large 4 still a preview?

It launched on October 6, 2026 as a public preview with weights promised by the end of the month. A later, undated Mistral post says it is releasing the weights of an improved version with about three more weeks of reinforcement learning, and Medium and High reasoning modes alongside Low. Mistral reports its own Terminal-Bench 4.0 score rising 13 percentage points to 41%; we have not tested that.

Is this the official Mistral website?

No. This is an unofficial page, not affiliated with or endorsed by Mistral AI. Facts link to Mistral’s own pages where possible.