Mistral published Mistral Large 4 on October 6, 2026 as a public preview, and the price on the page today is not the price buyers will pay for long. Mistral's pricing page lists an original price of $1.36 per 1M input tokens and $4.18 per 1M output tokens, with a sale price of $0.68 and $2.09 during launch. That is a 50% promotion, and when it ends the bill doubles without a single token changing.
This page covers the rates, when the promotion ends, who sells the model today, and what a buyer can do before the price resets. For how it compares with other models, see the LLM API pricing comparison.
Mistral Large 4 official API pricing
Rates from Mistral's pricing page and model card, standard tier, USD per 1M tokens:
| Token type | Launch sale price | Original price |
|---|---|---|
| Input | $0.68 | $1.36 |
| Cached input | $0.07 | $0.14 |
| Output | $2.09 | $4.18 |
The model name in the API is mistral-large-4. Mistral's model card lists a 1M token context window. OpenRouter lists 524K context with up to 256K output tokens, so check the limit on the route you use.
When the launch promotion ends
Mistral's own pages mark the rate as a sale price but do not state an end date. Vercel's AI Gateway listing for the same model says promotional pricing ends on October 20, 2026. Treat that as the working date and confirm it on Mistral's pricing page before you budget around it.
The cost of the reset is easy to size. For a workload of 10M input and 1M output tokens per day:
| Rate | Daily | Annual |
|---|---|---|
| Launch sale price | $8.89 | $3,245 |
| Original price | $17.78 | $6,490 |
Same workload, same model, $3,245 a year apart, and the date of the change is set by the seller. A team that builds on Mistral Large 4 this week at $0.68 is planning around a price that is scheduled to go away.
This is the case where a forward contract earns its place. A launch promotion is a temporary price. A forward is how a buyer locks future capacity at a known price before the rate goes back up. That is what the Mercatus Forward Market is built for, and how to lock in token prices walks through the mechanics.
One host today, open weights on October 31
Right now there is one seller. OpenRouter lists a single provider for Mistral Large 4, Mistral itself, at the same $0.68 and $2.09 sale rate, with a median throughput of 55 tokens per second. Vercel's gateway routes to Mistral as well. Mistral says the preview runs on its own datacenters in Europe.
Mistral has said it will release the weights by the end of October, and its Hugging Face page shows October 31, 2026 as the current expected date. Until then no third party can host the model, so there is no second price to compare against.
Put the two dates together. If the promotion ends October 20 and the weights arrive October 31, there is a stretch where the model costs full price and only one company sells it. After the weights are out, other hosts can list it and a price spread will form, the same way it did for other open-weight models. Why token prices differ explains where that spread comes from.
How the list price compares
At the original price, Mistral Large 4 sits in the same band as other large open-weight models. All rates per 1M tokens:
| Model | Input | Output | Source |
|---|---|---|---|
| Mistral Large 4 (launch sale) | $0.68 | $2.09 | Mistral |
| Mistral Large 4 (original) | $1.36 | $4.18 | Mistral |
| Mistral Large 3 | $0.50 | $1.50 | Mistral |
| GLM 5.3, hosted by Mistral | $1.40 | $4.40 | Mistral |
| DeepSeek V4 Pro, off-peak | $0.66 | $1.98 | DeepSeek |
| DeepSeek V4 Pro, peak | $1.32 | $3.96 | DeepSeek |
Two things stand out. The launch sale price lands almost exactly on DeepSeek V4 Pro's off-peak rate, and the original price lands almost exactly on its peak rate. And compared with Mistral Large 3, the original price is about 2.7 times higher on input and 2.8 times higher on output, so this is a new and larger model at a higher price point, not a like for like replacement.
For the DeepSeek side of that table, see DeepSeek V4 Pro API pricing.
What Mistral Large 4 is
From Mistral's model card and announcement:
- Size: 1.05 trillion total parameters with 52 billion active (49 billion per token before embeddings and output layers), plus a 1.6 billion parameter vision encoder.
- Architecture: granular Mixture of Experts, multimodal input.
- Status: public preview, version 26.10.
- Training: 3,800 NVIDIA Grace Blackwell GPUs in Mistral's own European datacenters.
- Weights: described as open-weight, not yet downloadable. The license for the weights has not been published.
What buyers should do before the promotion ends
Budget at the original price. Any cost model for Mistral Large 4 should use $1.36 and $4.18. The sale rate is a trial window.
Use the window to test. Half price for two weeks is a cheap way to run evals and find out whether the model earns the full rate on your workload.
Watch for the second host. Once the weights are out, compare listings before committing volume. The official API is not always the cheapest seller of an open-weight model.
Lock what you can. If the model fits and the volume is predictable, a forward contract fixes the price for the months ahead so the reset happens on your terms.
To track token prices across models and hosts, see the Mercatus Token Index.
Frequently asked questions
How much does the Mistral Large 4 API cost?
During the launch promotion, $0.68 per 1M input tokens, $0.07 per 1M cached input tokens, and $2.09 per 1M output tokens. Mistral lists the original price as $1.36, $0.14, and $4.18.
When does the Mistral Large 4 launch price end?
Mistral's pricing page does not give a date. Vercel's AI Gateway listing says promotional pricing ends on October 20, 2026.
Is Mistral Large 4 open weight?
Mistral describes it as open-weight and says the weights will be released by the end of October 2026. Its Hugging Face page shows October 31, 2026 as the current expected date. The weights were not available as of October 7, 2026.
Who hosts Mistral Large 4?
As of October 7, 2026, only Mistral. OpenRouter lists one provider and Vercel's gateway routes to Mistral.
Is Mistral Large 4 more expensive than Mistral Large 3?
Yes. Mistral Large 3 is $0.50 input and $1.50 output per 1M tokens. Mistral Large 4's original price is $1.36 and $4.18.
Methodology
Rates, model name, parameter counts, and context window from Mistral's API pricing page and Mistral Large 4 model card. Launch date, weights timing, and training details from Mistral's announcement of October 6, 2026. Expected weights date from Mistral's Hugging Face page. Host count and throughput from OpenRouter. Promotion end date from Vercel's AI Gateway listing. DeepSeek rates from DeepSeek's API pricing page. Workload examples use stated token volumes at 365 days; recompute with your own mix. Last verified: 2026-10-07.