Mistral Small 4 API pricing and cost calculator
Short answer: Mistral Small 4 costs $0.15 per 1M input tokens and $0.6 per 1M output tokens on Mistral's standard tier (checked 2026-11-15), with no cached-input rate listed. A 2,000-input / 500-output chat call costs $0.0006, so 1,000 calls cost $0.6000. Output is 4.0× the input price, which is why long answers dominate a bill.
Nearest-priced models, this workload
Showing the 9 models closest in price; change any input to rank every model.
| Model | Per call | Per 1K calls |
|---|---|---|
| GPT-6 Luna | $0.00045 | $0.4500 |
| Claude Haiku 5.5 | $0.00045 | $0.4500 |
| Ministral 3 14B | $0.0005 | $0.5000 |
| GPT-4o mini (deprecated) | $0.0006 | $0.6000 |
| Mistral Small 4 | $0.0006 | $0.6000 |
| GPT-5.6 Luna | $0.0010 | $1.00 |
| GPT-5.4 nano (deprecated, retires 2027-04-01) | $0.00103 | $1.03 |
| Codestral 25.08 | $0.00105 | $1.05 |
| DeepSeek Flash (V4.1) | $0.0012 | $1.20 |
o4-mini is shut down; its last published list rate is kept for comparing old invoices, not as a model you can call today — the model page names its replacement.
At a glance
| API model ID | mistral-small-2603 |
| Context window | 256K tokens |
| Inputs | text, image |
Every price tier
| Tier | Price per 1M tokens |
|---|---|
| Standard input | $0.15 |
| Cached input (cache read) | – |
| Output (incl. reasoning tokens) | $0.6 |
| Batch API input / output | $0.075 / $0.3 (50% off) |
Source: Mistral pricing and the Mistral Small 4 model page, checked 2026-11-15.
API name mistral-small-2603; 256K context; batch −50%; Apache 2.0
Worked examples
| Workload | Per call | Per 1K calls | Per 1K, Batch |
|---|---|---|---|
| Chat reply · 1K in / 500 out | $0.00045 | $0.4500 | $0.2250 |
| RAG answer · 10K in / 1K out | $0.0021 | $2.10 | $1.05 |
| RAG with 80% cached prefix | $0.0021 | $2.10 | $1.05 |
| Long document · 100K in / 2K out | $0.0162 | $16.20 | $8.10 |
| Whole-codebase prompt · 400K in / 4K out | exceeds the 256K context window | ||
When to pick it
- Mistral describes it as: “Hybrid model unifying instruct, reasoning and coding in a single efficient model.”
- Cheapest current model from another provider with at least a 256K context window: GPT-6 Luna at $1.50 per 1,000 RAG calls.
- Output costs 4× input here, so long prompts matter about as much as long answers.
- Background jobs (classification, nightly summaries, evals) can go through the Batch API for 50% off.
Where it sits in the Mistral lineup
Among the 6 Mistral models we track, Mistral Small 4 ranks #3 from cheapest at a 10K-in / 1K-out workload ($2.10 per 1,000 calls). The next cheaper option is Ministral 3 8B at $1.65 (0.79× the cost, 256K context). Stepping up to Ministral 3 14B costs $2.20 (1×). The cheapest in the lineup, Ministral 3 3B, is 1.91× cheaper.
Performance benchmarks
Measured scores copied from the dated leaderboard pages linked in the table: SWE-bench Verified, LiveCodeBench, GPQA Diamond, AIME, Humanity's Last Exam, MMMU-Pro. A benchmark this model does not publish is left out rather than estimated, and scores from different benchmarks are never mixed into one number.
On SWE-bench Verified, Mistral Small 4 ranks #2 of the 3 Mistral models we track that publish this benchmark.
| Benchmark | Score | Rank in the Mistral lineup | Cost per quality point | Source, checked |
|---|---|---|---|---|
| SWE-bench Verified | 60.1% | #2 of 3 | $0.00437 / point | trustthebench.com · |
| LiveCodeBench | 63.1% | #2 of 3 | $0.00416 / point | trustthebench.com · |
| GPQA Diamond | 57.1% | #2 of 4 | $0.0046 / point | artificialanalysis.ai · |
| AIME | 71.5% | #2 of 3 | $0.00367 / point | trustthebench.com · |
| Humanity's Last Exam | 3.8% | #4 of 4 | $0.0691 / point | artificialanalysis.ai · |
| MMMU-Pro | 46.5% | #2 of 4 | $0.00565 / point | artificialanalysis.ai · |
GPQA/MMMU-Pro/HLE from Artificial Analysis entry 'Mistral Small 4 (Non-reasoning)'.
Cost per quality point = blended standard price (3 input tokens : 1 output token per 1 million) ÷ the benchmark score: what one point of measured quality costs at list price, not a task cost. Ranks compare only models from the same provider on the same benchmark. All benchmarks ranked by price and performance · Best LLM for coding · Compare two models
Related
- All Mistral models and prices
- Benchmarks: price vs performance
- Best LLM for coding in 2026
- Find a model for your use case
- Mistral cost calculator
- Prompt caching guide
- Batch API guide
- How to price an AI SaaS from measured usage
- Reasoning tokens: why thinking models cost more
Ready to build? Create an API key in the official Mistral console.
Questions
How much does Mistral Small 4 cost per 1M tokens?
$0.15 for input, $0.6 for output and – for cached input on the standard tier. Source: the official Mistral pricing page, checked 2026-11-15.
How much do 1,000 Mistral Small 4 API calls cost?
At 2,000 input and 500 output tokens per call, one call costs $0.0006 and 1,000 calls cost $0.6000 at list price. Batch and caching are not included; both lower the total.
Is Mistral Small 4 expensive compared with other models?
On a 2,000-input / 500-output chat call it ranks #8 of 55 verified models by cost. The cheapest is Ministral 3 3B at $0.00025 per call.
Does Mistral Small 4 publish a cached-input rate or a batch discount?
Its official pricing page lists no cached-input rate on its official page, so cached tokens are billed at the normal input rate and a Batch API multiplier of 0.5, a 50% discount. Caching pays off when a system prompt or document prefix repeats across calls; batch applies only to asynchronous jobs nobody is waiting on. Both are priced on top of the standard rates above, never instead of them.
Next step → Estimate cost per user with Mistral Small 4
Price data last checked 2026-11-18 against official sources · how we verify
Price-change alerts
One short email when a provider changes a price or a promo ends. No spam, unsubscribe any time.
Tools, skills & plugins that use this model
Matched to this model's profile: cheaper models are paired with workflow and agent tools, million-token-context models with frameworks and open-source projects, and reasoning models with coding agents. Each entry links to its token-cost page or official site, and the per-task price is calculated on this model from the tool data rather than copied from another model.
- n8n AI Agent — Workflow automation whose AI Agent nodes resend tool results every step. ($0.00825 per task, Measured)
- Zapier Agents — Zapier's AI agents, billed per activity instead of per token. ($0.0042 per task, Estimated)
- Claude Code — Anthropic's terminal coding agent for multi-step software tasks. ($0.0114 per task, Documented)
- OpenAI Codex CLI — OpenAI's open-source terminal coding agent, billed via ChatGPT credits. ($0.0096 per task, Documented)