Mistral La

mistral.ai · Compute, Inference · 3 records · JSON

Mistral La Plateforme Computecaveatany-phoneno cardphoneDrequest-rate
Mistral La Plateforme Inferencecaveatemailno cardDrequest-rate
Mistral Compute / AI Cloud Computedeademail?card?Rwrite-gated

Mistral La Plateforme Compute

caveatany-phoneno cardphoneDrequest-rate

docs.mistral.ai/admin/billing-usage/usage-limits · inference API · compute/mistral-la-plateforme · JSON · compare

Free tier	Free mode still exists and its limits are now unpublished. The docs say 'Free mode lets you create API keys and use included monthly usage within the limits shown on the Limits page', and the help centre names the three limit dimensions, 'Requests per second (RPS)', 'Tokens per minute' and 'Tokens per month'. No figure appears on any public Mistral page: the docs point at admin.mistral.ai/plateforme/limits, which is behind a login. The 'USD 10 /mo in API credits' line that third parties still quote is no longer on mistral.ai/pricing, and a 2026-09-01 archive capture does not carry it either. unit none; metered on wall-clock; recurring-monthly-allowance
Price	none stated
Per-object cap	not applicable
Retention	not applicable
Egress	not stated
Identity	any-phone a one-time SMS verification on any phone number, and no card: the record's notes report this from third-party signup guides, and state that no Mistral page itself documents a phone step, so this is reported rather than documented. No authorisation amount is stated because no payment method is taken; debit and prepaid acceptance are therefore not applicable on the free mode. What the guides say the SMS is sent to is not recorded
Limits	request-rate the number that binds is not published, so the honest answer is that the monthly token cap is unknown until an account exists; third-party guides that still print '1 billion tokens a month' and '1 request a second' describe the retired Experiment plan, whose help pages now return 404. source the number that binds is not published, so the honest answer is that the monthly token cap is unknown until an account exists; third-party guides that still print '1 billion tokens a month' and '1 request a second' describe the retired Experiment plan, whose help pages now return 404.
Binding	request-rate the number that binds is not published, so the honest answer is that the monthly token cap is unknown until an account exists; third-party guides that still print '1 billion tokens a month' and '1 request a second' describe the retired Experiment plan, whose help pages now return 404.
Interface	rest inferred from the tooling the provider documents; not probed
Tooling	not stated
Composes with	an unpublished monthly cap is still a cap: the only safe composition is to treat the free key as a prototype lane and move any recurring job to Cloudflare's published 10,000 Neurons a day, which at least prints its meter.
Reliability	no reliability event recorded
Notes	Refutation of a widely repeated number: '1 billion tokens a month on Mistral's free tier' belongs to the Experiment plan that Mistral retired, and no current Mistral page states it. The 'USD 10 a month in API credits' line was on the pricing page in early October 2026 according to a change tracker; it is absent today and absent from the 2026-09-01 capture, so the exact date it left is unverified and the current state is that no figure is published. Identity: third-party signup guides report a one-time SMS verification and no card, and no Mistral page states a phone step; recorded as any-phone, reported rather than documented. Data on Free mode may be used for training unless opted out, which is a technical reliability fact a batch job should account for.
Evidence rung	documented
Checked	2026-10-22 (live); stale on 2026-11-21
Parsed numbers	`$.limits[0].value` limit:request-rate 1.0 request

Evidence (4)

date	source	what it proved
2026-10-22	docs.mistral.ai/admin/billing-usage/usage-limits	'Free mode lets you create API keys and use included monthly usage within the limits shown on the Limits page', quoted; no figure on the page
2026-10-22	help.mistral.ai/en/articles/698531-why-am-i-hitting-my-api-u	'The API enforces three limits: Requests per second (RPS)... Tokens per minute... Tokens per month', and 'See your current limits, usage, and tier on the Limits page', quoted
2026-10-22	mistral.ai/pricing	the Free plan card lists Vibe and Studio features only; no API credit figure appears anywhere on the page
2026-09-01	web.archive.org/web/20260901163208/https://mistral.ai/pricin	the 2026-09-01 capture also carries no 'USD 10 /mo in API credits' line, so the figure was already gone by September

More fields (1)

{
 "wave2": {
  "class": "refutation",
  "free_tier_kind": "recurring-monthly-allowance",
  "shape": {
   "vcpu": null,
   "ram_gib": null,
   "gpu_class": null,
   "gpu_count": null
  }
 }
}

Unknown or withheld values (8)

$.egress.metered egress-unknown
$.egress.terms not-stated
$.per_object_cap.notes process
$.per_object_cap.stated not-applicable
$.reliability_note none-observed
$.retention.notes process
$.retention.stated not-applicable
$.tooling not-stated

Mistral La Plateforme Inference

caveatemailno cardDrequest-rate

api.mistral.ai/v1 · inference endpoint, keyed free tier · inference/mistral-la-plateforme · JSON · compare

Free tier	a free API tier exists: 'We offer various tiers on the platform, including a free API tier with restrictive rate limits. The free API tier is designed to allow you to try and explore our API.' No number is published on that page unit requests; metered on request-count; the live page returns 404 today; the text quoted is from the Wayback capture of 2025-10-09T00:54:12Z, reached through the archive broker. The page states the numbers are per workspace and visible only at admin.mistral.ai/plateforme/limits, so the ceiling is behind a login
Price	none stated
Per-object cap	not applicable: an inference endpoint stores nothing for the caller
Retention	not applicable
Egress	the response is the payload; no separate egress term is published
Identity	email an account and an API key
Limits	request-rate Requests per second (RPS) and tokens per minute and per month, set at the workspace level and defined by usage tier; the free tier's values are published only at admin.mistral.ai/plateforme/limits after sign-in source no numeric limit was read. The docs site's tier page returned 404 on 2026-10-22, so the ceiling was not read; the read broker was not retried on an alternative path.
Binding	request-rate the per-workspace request and token rates bind, and the free tier's values are behind a login, so the ceiling cannot be sized before an account exists
Interface	rest
Tooling	any OpenAI-compatible client
Composes with	pairs with an agent after a signup; the scarce link is the account, and the free tier's ceiling is unread
Reliability	no reliability event recorded; no key was available
Notes	The archive shows it existed in October 2025 with the same content shape and the same deferral to a post-login limits page, so a move is the more likely and nothing suggests the free tier closed. The rung is documented, not measured: the 401 proves the endpoint and not the allowance. This record is stale by design: its only first-party evidence for the free tier is an archived page 378 days old, because the live documentation path 404s. The alternative was to keep the tier unrecorded, which would have left a claim with no evidence at all. Re-read it when the docs site serves a tier page again.
Evidence rung	documented the free tier's existence and its scope are from the archived first-party page, because the live path 404s; the 401 gate was probed today. No number is published in either place, so the ceiling is documented as existing and not as sized.
Checked	2026-10-23 (live); stale on 2026-11-22

Evidence (3)

date	source	what it proved
2026-10-22	docs.mistral.ai/deployment/laplateforme/tier/	returned 404 with a soft-404 body ('¿Meow? Page not found.'), through the raw fetch and the read broker alike. The documentation path has moved
2025-10-09	web.archive.org/web/20251009005412/https://docs.mistral.ai/d	the archived 'Rate limit and usage tiers' page states 'We offer various tiers on the platform, including a free API tier with restrictive rate limits', that limits are 'set at the workspace level' and defined by usage tier, and that the current values are visible only at admin.mistral.ai/plateforme/limits
2026-10-22	probe:https://api.mistral.ai/v1/chat/completions	an unauthenticated POST returned 401 with a 29-byte body, {"detail":"Invalid API Key"}. The endpoint exists and gates on a key

Unknown or withheld values (1)

$.retention.stated not-applicable

Mistral Compute / AI Cloud Compute

deademail?card?Rwrite-gated

mistral.ai/products/aicloud/ · GPU cloud · compute/mistral-compute · JSON · compare

Free tier	not stated unit none; metered on wall-clock
Price	none stated
Per-object cap	not stated
Retention	not stated
Egress	not stated
Identity	email (placeholder) not published
Limits	write-gated not published source not stated
Binding	write-gated
Interface	unknown
Tooling	not stated
Composes with	not stated
Reliability	the product page returns 404 and lands on a support shell
Notes	The product page is gone. The URL now lands on a help-centre shell that offers support channels and no product, price or signup, so whatever Mistral sold at /products/aicloud is no longer sold at that address.
Evidence rung	reported
Checked	2026-10-23 (inherited); stale on 2026-11-06

Evidence (2)

date	source	what it proved
2026-10-02	mistral.ai/products/aicloud/	not published
2026-10-22	mistral.ai/products/aicloud/	the origin returns 404 for mistral.ai/products/aicloud/ on every route, including ?__provider=firecrawl and the raw fetch, on 2026-10-22; the capture is Mistral's help-centre page, which is where the 404 lands

More fields (13)

{
 "talaria_m10h_usd": null,
 "talaria_m24h_usd": null,
 "nemo_m10h_usd": null,
 "keep_rate": null,
 "keep_rate_basis": "no field in the source",
 "billing_model": "unclassified",
 "shape": null,
 "session_cap": null,
 "persistent_storage": null,
 "network_out": null,
 "price_note": null,
 "free_tier_kind": "none",
 "reclassified_from": "dead"
}

Unknown or withheld values (19)

$.chaining not-stated
$.confidence_note withheld
$.egress.metered egress-unknown
$.egress.terms not-stated
$.evidence[0].proved process
$.extras.price_note process
$.free_tier.notes withheld
$.free_tier.stated not-stated
$.identity_cost.detail defaulted
$.identity_cost.kind identity-defaulted
$.interface interface-unknown
$.interface_detail withheld
$.limits[0].value withheld
$.limits_note not-stated
$.per_object_cap.notes process
$.per_object_cap.stated not-stated
$.retention.notes not-extracted
$.retention.stated not-stated
$.tooling not-stated

Corrections

2026-10-22 Mistral's AI-cloud page is a 404, and the record sold it as paid capacity