API model string
mistral-medium-3.5Context window
128K tokens
Max output not published
- Knowledge cutoff
- not published
- Released
- 2025–2026
- Modalities
- text in, text out
Last verified July 10, 2026
Rate limits by tier
Mistral La Plateforme uses three access tiers: a free Experiment tier for evaluation, a pay-as-you-go tier unlocked by adding a billing card, and a custom Enterprise tier. Exact RPM numbers are no longer published publicly — your live org ceilings appear in Admin Console → Limits. The slug mistral-medium-3 redirects to this page; all data below describes the current Mistral Medium 3.5 endpoint.
| Tier | Requirements | RPM | TPM | RPD | Concurrent | Notes |
|---|---|---|---|---|---|---|
| Free (Experiment) | No billing card required — sign up at mistral.ai and call the API immediately for evaluation | not published (check Admin Console → Limits) | ~1B tokens/month cap | not published | — | Rate-limited and intended for evaluation only. Not suitable for production workloads. The -latest alias may silently upgrade to Medium 3.5 if you were on Medium 3 — check your model string. |
| Pay-as-you-go (paid) | Enable billing in La Plateforme account settings | ~300 RPM baseline (per third-party trackers — confirm in Admin Console) | not published | not published | — | No monthly minimums. Mid-tier between Mistral Small and Mistral Large. Limits grow with usage history. Pin 'mistral-medium-3.5' explicitly — safer than 'mistral-medium-latest' given the Medium 3 → 3.5 transition. |
| Enterprise (custom) | Contact Mistral sales | not published | not published | not published | — | Includes SAML SSO, audit logs, custom limits, and dedicated SLA. |
Swipe the table sideways to see every limit column.
- 1.Mistral no longer publishes exact RPM numbers publicly; Admin Console → Limits is the authoritative source for your organization.
- 2.Version-to-price mapping for Medium is actively changing: Medium 3 was $0.40/$2.00; Medium 3.5 is cited at ~$1.50/$7.50 on some third-party trackers. Verify on mistral.ai/pricing before quoting either figure.
- 3.Pinning 'mistral-medium-3.5' is safer than 'mistral-medium-latest' — the -latest alias can silently switch to a new version mid-deploy with different pricing.
Limits verified against the Mistral AI docs, July 10, 2026.
Token pricing
What you pay per million tokens (USD). Input and output are billed separately.
Input
$0.40
per 1M tokens
Output
$2.00
per 1M tokens
Batch discount
−50%
on batch jobs
- The $0.40/$2.00 figures reflect the Mistral Medium 3 rate card from third-party trackers and are used here as a conservative floor. Mistral Medium 3.5 is listed at approximately $1.50 in / $7.50 out on some trackers — pricing has shifted with the version upgrade. Verify the current rate on mistral.ai/pricing before populating any budget calculation.
- Context caching is available and discounts repeated-prefix input; exact cached rate not published — verify on mistral.ai/pricing.
- Batch API provides 50% off input and output for non-latency-sensitive workloads. Confirm batch support for the mistral-medium-3.5 endpoint before building the pipeline.
- Reasoning/thinking tokens (applicable to Magistral models) are billed at the output rate — this does not directly apply to Mistral Medium.
Side-project chatbot
$4.00
per month
Assumptions
5M tokens in / 1M tokens out per month (Medium 3 rate $0.40/$2.00)
5M × $0.40/MTok + 1M × $2.00/MTok = $2.00 + $2.00 = $4.00 (verify Medium 3.5 rate — could be up to $15 at $1.50/$7.50)
Growing SaaS
$48
per month
Assumptions
60M tokens in / 12M tokens out per month (Medium 3 rate)
60M × $0.40/MTok + 12M × $2.00/MTok = $24 + $24 = $48 (no published cached rate — skip cache scenario; verify Medium 3.5 rate)
High-volume pipeline
$320
per month
Assumptions
400M tokens in / 80M tokens out per month (Medium 3 rate)
400M × $0.40/MTok + 80M × $2.00/MTok = $160 + $160 = $320 (at Medium 3.5 rate of $1.50/$7.50 this would be ~$600 + $600 = $1,200 — verify)
Run your own numbers
Drag your real monthly token volumes and watch the bill update live — priced against rival models at the same usage.
Estimated mistral-medium-3.5 spend
$22/mo
Input: $12
Output: $10
30M in × $0.400 + 5M out × $2.00 = $22
Same volume, priced across models
- Gemini 2.5 FlashCheapest$22
- mistral-medium-3.5This model$22
- Claude Haiku 4.5$55
- Mistral Large 3$90
Rivals priced at their published input/output rates for the same monthly volumes. Prompt caching is model-specific, so it is applied to mistral-medium-3.5 only. Estimates for comparison; real bills vary with request shape and long-context surcharges.
mistral-medium-3.5 vs the alternatives
Mistral Medium 3.5 competes in the mid-tier segment against Mistral's own flagship, Gemini 2.5 Flash, and Claude Haiku 4.5 — positioning on EU data residency and price when Medium 3 rates apply.
| Aspect | mistral-medium-3.5 | Mistral Large 3 | Gemini 2.5 Flash | Claude Haiku 4.5 |
|---|---|---|---|---|
| Context window | 128K | 128K | 1M | 200K |
| Input $/MTok | $0.40–1.50 (verify) | $2.00 (verify) | $0.30 | $1.00 |
| Output $/MTok | $2.00–7.50 (verify) | $6.00 (verify) | $2.50 | $5.00 |
| EU data residency | Yes (Paris) | Yes (Paris) | No (Google infra) | No (US-only option) |
| Batch discount | 50% | 50% | 50% | 50% |
| Free eval tier | Yes (Experiment) | Yes (Experiment) | Yes (limited) | No |
| Open weights | No | No | No | No |
| Best for | EU mid-complexity tasks at competitive price (if $0.40) | flagship quality | speed + 1M context | Anthropic ecosystem |
Swipe the table sideways to see every model.
Hitting a 429? The playbook
The exact errors you'll see
HTTP 429 Too Many Requestsrate_limit_errorRetry-After: <seconds>Why it happens & how to fix it
Free Experiment tier RPM exhausted
Switch to pay-as-you-go — the free tier is intentionally constrained for evaluation. Add a billing card in La Plateforme account settings to immediately unlock the paid baseline (~300 RPM).
Paid ~300 RPM baseline exceeded during burst
Implement exponential backoff with Retry-After. Space requests across time or use the Batch API for non-latency-sensitive workloads. Check Admin Console → Limits for self-serve increase options.
-latest alias silently upgraded to Medium 3.5 with different pricing
Pin the explicit model version string 'mistral-medium-3.5' in your API calls. The -latest alias can switch to a new version mid-deploy, changing both rate behavior and pricing unexpectedly.
Monthly ~1B token cap on free tier exhausted
Add a billing card to upgrade to pay-as-you-go; the paid tier has no monthly token cap.
Medium 3 vs Medium 3.5 endpoint confusion causing wrong pricing expectations
Confirm your model string in API calls and check mistral.ai/pricing for the current version mapping. Medium 3 ($0.40/$2.00) and Medium 3.5 (~$1.50/$7.50) have materially different rates on some trackers.
Retry strategy
Standard HTTP 429 returns a Retry-After header specifying wait seconds — honor it precisely. Use exponential backoff as a fallback: delay = min(2^attempt, 60) seconds ± 10% jitter to prevent thundering herd. La Plateforme is OpenAI-compatible; initialize the OpenAI SDK with base_url='https://api.mistral.ai/v1' and substitute your Mistral API key.
1import OpenAI from 'openai';23const client = new OpenAI({4 apiKey: process.env.MISTRAL_API_KEY,5 baseURL: 'https://api.mistral.ai/v1',6});78async function callMistralMediumWithRetry(9 prompt: string,10 maxRetries = 511): Promise<string> {12 for (let attempt = 0; attempt < maxRetries; attempt++) {13 try {14 const response = await client.chat.completions.create({15 model: 'mistral-medium-3.5', // pin explicit version, not mistral-medium-latest16 messages: [{ role: 'user', content: prompt }],17 });18 return response.choices[0].message.content ?? '';19 } catch (err: unknown) {20 const error = err as { status?: number; headers?: Record<string, string>; message?: string };21 if (error.status === 429) {22 const retryAfter = parseFloat(error.headers?.['retry-after'] ?? '0');23 const jitter = 1 + (Math.random() * 0.2 - 0.1);24 const delay = retryAfter > 025 ? retryAfter * 100026 : Math.min(Math.pow(2, attempt) * 1000 * jitter, 60000);27 console.warn(`Rate limited. Retrying in ${(delay / 1000).toFixed(1)}s (attempt ${attempt + 1}/${maxRetries})`);28 await new Promise(resolve => setTimeout(resolve, delay));29 } else {30 throw err;31 }32 }33 }34 throw new Error('Max retries exceeded for Mistral Medium API call');35}How to raise your limits
The ladder from the starter tier to enterprise — what each rung takes, and what it unlocks.
Experiment (free)
ImmediateSign up at mistral.ai — no card required. Access the Mistral Medium endpoint for evaluation immediately.
Unlocks: ~1B tokens/month cap, low RPM (check Admin Console → Limits), evaluation use only
Pay-as-you-go
Immediate upon card additionAdd a billing card in La Plateforme account settings. No minimum monthly spend required.
Unlocks: ~300 RPM baseline (per third-party trackers — verify in Admin Console), no monthly token cap, production-ready access
Admin Console → Limits (self-serve)
Immediate for self-serve increasesNavigate to Admin Console → Limits to view your current org ceilings. Some increases are available without contacting support.
Unlocks: Higher org-level limits without a sales conversation
Enterprise
Sales negotiationContact Mistral sales for custom limits, SAML SSO, audit logs, and a dedicated SLA.
Unlocks: Custom RPM/TPM, dedicated support, enterprise compliance features
Cut your token spend
Context caching
Discount on repeated-prefix input (exact rate not published — verify on mistral.ai/pricing)Structure prompts with a stable shared system-prompt prefix (product docs, persona, instructions) that remains constant across requests. Mistral caches repeated prefixes and applies a discount to cached input tokens.
Batch API
-50% on input and output cost for async workloadsSubmit classification, summarization, or data extraction tasks as batches. Check that mistral-medium-3.5 supports batch submissions in your Admin Console before building the pipeline.
Pin explicit model version
Prevents silent pricing and behavior changes between Medium 3 and 3.5Use 'mistral-medium-3.5' as your model string, not 'mistral-medium-latest'. This is especially important given the Medium 3 → 3.5 transition and the significant pricing difference between versions.
Downgrade to Mistral Small for simpler tasks
Significant cost savings — Mistral Small is cheaper and open-weightRoute simple classification, intent detection, or low-stakes summarization to Mistral Small. Reserve Medium for tasks that need meaningfully more reasoning capacity than Small provides.
EU compliance at no premium
GDPR-adjacent requirements met with zero additional costMistral routes all La Plateforme traffic through Paris by default. No regional surcharge, no additional configuration — unlike AWS/Azure residency options that typically add cost.
Monitor mistral.ai/pricing monthly
Avoids unexpected billing surprises during active version transitionsMedium pricing is actively changing (Medium 3 → 3.5 transition). Set a calendar reminder to re-verify the rate card on mistral.ai/pricing — the $0.40/$2.00 vs $1.50/$7.50 discrepancy is material for any budget calculation.
Frequently asked questions
How do I increase Mistral Medium rate limits?
Add a billing card to your La Plateforme account to immediately upgrade from the free Experiment tier to the pay-as-you-go baseline (~300 RPM per third-party trackers). Then check Admin Console → Limits for your current org ceilings — some self-serve increases are available without contacting support. For higher custom limits, contact Mistral sales.
Is Mistral Medium API free?
Yes — Mistral's free Experiment tier requires no billing card and includes approximately 1 billion tokens per month. The RPM ceiling is low (not published publicly; check Admin Console → Limits) and the tier is intended for evaluation only. Paid pay-as-you-go access has no monthly minimum and unlocks the production baseline.
What is Mistral Medium 3.5 pricing?
Pricing has shifted between versions. Mistral Medium 3 was $0.40 input / $2.00 output per million tokens; Mistral Medium 3.5 is cited at approximately $1.50/$7.50 on some third-party trackers. Both figures need verification on mistral.ai/pricing before use in any budget calculation. A 50% batch discount is available, and context caching discounts repeated-prefix input at a rate not yet published.
Mistral Medium vs Mistral Large — which should I use?
Use Mistral Medium 3.5 for mid-complexity tasks where you need a balance of quality and cost, especially in EU-regulated environments. Mistral Large 3 offers higher reasoning quality at $2.00/$6.00 per MTok (verify) and is appropriate when output quality is the primary constraint. If Medium 3's $0.40/$2.00 rate applies to 3.5, Medium is roughly 5× cheaper on output.
What is the difference between Mistral Medium 3 and Mistral Medium 3.5?
Mistral Medium 3.5 is a version upgrade of Mistral Medium 3 within the same product line — not a full model replacement. The slug mistral-medium-3 redirects to this page. Pricing may differ significantly between versions ($0.40/$2.00 for Medium 3 vs ~$1.50/$7.50 for 3.5 on some trackers). Medium 3 may still be callable via a pinned version string — verify on mistral.ai/pricing.
Does Mistral Medium support EU data residency?
Yes. Mistral routes all La Plateforme traffic through Paris, France by default. This provides EU data residency without additional cost or configuration — unlike AWS or Azure residency options that typically add a regional surcharge. This applies to Mistral Medium 3.5 and all other La Plateforme models.
How do I fix Mistral Medium 429 errors in production?
Read the Retry-After header in the 429 response and wait that many seconds. As a fallback, use exponential backoff: delay = min(2^attempt, 60) seconds ± 10% jitter. If you're on the free Experiment tier, add billing to access the paid baseline. If you're hitting the paid ~300 RPM limit during bursts, use the Batch API for non-latency-sensitive work or contact Mistral for a limit increase via Admin Console → Limits. RapidDev can help scope retry architecture for your specific throughput requirements — rapidevelopers.com/contact.
Can I still use the mistral-medium-latest model string?
Technically yes, but it carries version-drift risk. The -latest alias can silently upgrade from Medium 3 to Medium 3.5 mid-deploy, changing pricing and potentially behavior. Pin the explicit string 'mistral-medium-3.5' in production to lock both the rate card and the model behavior. Check mistral.ai/pricing after any intentional version update.
We build AI apps that don't hit rate limits
- Retry, backoff & caching built in
- Multi-provider fallback routing
- Fixed price, you own the code
30-min call. No commitment.