No more mistral free tier?

[hakergeniusz]

Most API request also fail from the API, except Ministral and Codestral models:

Mistral Medium and Mistral Small models return 429. On this Reddit post people also confirmed it. Only from playground Mistral Medium seems to work.

[diegosouzapw]

@hakergeniusz Thank you for the screenshots and the details. What they show, and what I can and can't say:

• The Mistral console banner says: "Your organization is on the Free plan. Upgrade to activate your API keys and access the full API", with New key greyed out. That is Mistral's own wording, on your account.
• The model list shows the Ministral, Codestral and Voxtral models passing the health check, and the Mistral Medium, Small and Magistral models failing. That matches what you describe: 429 on Medium/Small, and only Ministral/Codestral answering.

This conflicts with what our free-tier catalog says. We recorded Mistral's free mode as verified in the account console on 2026-09-02, with API keys allowed and a 1B-token monthly cap. I haven't re-checked it since, so I can't tell you whether Mistral changed the free plan, is rolling it out gradually, or whether this is specific to your organization. I also haven't confirmed the Reddit report you linked.

Two things would help: the exact body of one of the 429 responses (the message text, with your key hidden), and whether a second account in another organization shows the same banner. In the meantime I am tracking this in #16279 so the free-tier docs and the catalog get re-checked against Mistral's current terms; if it is confirmed, the "free" label and the Mistral share of our free-token headline need to change.

[hakergeniusz]

As you asked:

[hakergeniusz@acer-archlinux ~]% curl https://api.mistral.ai/v1/chat/completions \
  -H "Authorization: Bearer $MISTRAL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mistral-small-latest",
    "messages": [{"role": "user", "content": "Hello, Mistral!"}]
  }'
{"object":"error","message":"Rate limit exceeded","type":"rate_limited","param":null,"code":"1300","raw_status_code":429}%

[hakergeniusz@acer-archlinux ~]% curl https://api.mistral.ai/v1/chat/completions \
  -H "Authorization: Bearer $MISTRAL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "ministral-8b-2512",
    "messages": [{"role": "user", "content": "Hello, Mistral!"}]
  }'
{"id":"8cb2f266ff944a7882139455e667b1d8","created":1791697821,"model":"ministral-8b-2512","usage":{"prompt_tokens":8,"total_tokens":49,"completion_tokens":41,"prompt_tokens_details":{"cached_tokens":0},"service_tier":"standard"},"object":"chat.completion","choices":[{"index":0,"finish_reason":"stop","message":{"role":"assistant","tool_calls":null,"content":"Hello! 😊 How can I assist you today? Whether you have questions, need help with something, or just want to chat, I'm here for it.\n\nWhat's on your mind?"}}]}%

[hakergeniusz@acer-archlinux ~]% curl https://api.mistral.ai/v1/chat/completions -H "Authorization: Bearer $MISTRAL_API_KEY" -H "Content-Type: application/json" -d '{
    "model": "mistral-large-latest",
    "messages": [{"role": "user", "content": "Hello, Mistral!"}]
  }'
{"object":"error","message":"This model is not available in your subscription tier","type":"tier_not_allowed","param":null,"code":"1910","raw_status_code":403}%
[hakergeniusz@acer-archlinux ~]%

On a key created ~2 months ago. I barely called Mistral (probably not at all), and the rate limit probably is set to 0.

[hakergeniusz]

https://docs.mistral.ai/getting-started/quickstarts/studio/activate-and-generate-api-key

Docs still don't mention it (as of 11-10-2026 5:59AM UTC)

[PrinceThummar011]

Hi @hakergeniusz,

Thank you for sharing the screenshots of the API Keys page and the Playground!

To help us update OmniRoute's free-tier catalog accurately, could you also share a screenshot or details of your organization's Limits page (under console.mistral.ai → Limits or your Admin settings)?

Previously, OmniRoute recorded a 1,000,000,000 monthly token cap there. Seeing what is currently listed on your Limits page (e.g. if tokens/month is now set to 0, or missing entirely) will provide the missing evidence needed to update the catalog. (Please remember to mask any personal or account identifiers). Thanks!

[hakergeniusz]

[diegosouzapw]

@hakergeniusz Thank you, and thanks to @PrinceThummar011 for asking for the Limits page. I read your three screenshots. What they show:

• Limits page: per-model rate limits only (tokens per minute and requests per second: for example mistral-medium-latest, mistral-small-2603 and mistral-large-4 at 20,000 tokens per minute, ministral-8b at 625,000). In the part of the page you captured I don't see a monthly token cap anywhere. I can't tell from the screenshots whether there's one further down, or whether it was removed.
• Usage page: 25 API requests this month, "Total Estimated Cost $0.00". mistral-medium-latest (4 requests), ministral-8b and codestral-latest are marked Billable; the Vibe usage is marked "Not billable".

I can't draw a conclusion from this alone. The 1B-token monthly cap in our catalog was read from this Limits page on 2026-09-02, and the per-minute cards look like the ones that were there then, but I haven't compared them line by line. Could you scroll the Limits page to the top and check if there's a card or line named "Tokens per month" (or similar), and tell us the exact error body of one failing mistral-medium-latest call (the message field)? Until then I'm keeping #16279 open and not changing the free-tier numbers.

[hakergeniusz]

[hakergeniusz@acer-archlinux ~]% curl -sS -D /tmp/mistral-headers.txt https://api.mistral.ai/v1/chat/completions \
> -H "Authorization: Bearer ${MISTRAL_API_KEY}" \
> -H 'Content-Type: application/json' \
> -d '{"model":"mistral-medium-latest","messages":[{"role":"user","content":"Reply with exactly: test"}]}'
{"object":"error","message":"Rate limit exceeded","type":"rate_limited","param":null,"code":"1300","raw_status_code":429}                                                                                                                                     
[hakergeniusz@acer-archlinux ~]%

You meant this or something else?

[diegosouzapw]

@hakergeniusz Yes, that's exactly what I meant, thank you. It shows something useful: a single tiny request straight to Mistral's API, with no OmniRoute involved, gets 429 Rate limit exceeded (code 1300) on mistral-medium-latest. So the 429 comes from Mistral, not from OmniRoute, and OmniRoute can't fix it.

What I can't tell from it is the reason. A one-line request exceeding a limit that the Limits page shows as 20,000 tokens per minute suggests that, for this organization, that model's allowance is effectively zero or shared, but that's my reading and not something Mistral says. Your command saved the response headers to /tmp/mistral-headers.txt: if there are any lines starting with x-ratelimit or retry-after in it, please paste those (nothing else from that file, since it can contain your account details). Those headers would say which limit was hit.

I'm keeping #16279 open and leaving the catalog as it is until we know whether this is a plan change or specific to your organization.