Magistral Small API Pricing: Mistral's Budget Reasoning Model at $0.50/$1.50

Q: How much does Magistral Small cost?

Magistral Small costs $0.50 per million input tokens and $1.50 per million output tokens. The output-to-input ratio is 3:1 — lower than most models, which means output-heavy reasoning chains are relatively affordable compared to models like Claude Sonnet 5 ($2/$10) or GPT-5.4 ($2.50/$15).

Q: What is the context window of Magistral Small?

Magistral Small has a 256K token context window — double the 128K offered by Mistral Small 4 and most budget models. This makes it well-suited for complex reasoning tasks that require large context, such as multi-document analysis, long codebases, or detailed problem statements.

Magistral Small brings chain-of-thought reasoning to budget pricing — $0.50/M input, $1.50/M output with 256K context. The affordable reasoning option from Europe's leading AI lab.

Updated Aug 14, 2026 · 94 models tracked across 11 providers

TL;DR

Price: $0.50/M input, $1.50/M output — 3:1 output ratio, great for reasoning chains
Context: 256K tokens — double most budget models, handles complex multi-document reasoning
Provider: Mistral (Paris, France) — European data sovereignty
Key feature: Chain-of-thought reasoning — thinks through problems step by step
Best for: Math, logic, code reasoning, structured extraction, complex analysis
Trade-off: More expensive than non-reasoning models (Mistral Small 4 at $0.15/$0.60), but much better at tasks requiring systematic thinking

European AI: Mistral is headquartered in Paris with European API infrastructure. Magistral Small processes data on European data centers — ideal for EU companies needing GDPR-compliant AI reasoning without data leaving European jurisdiction.

Magistral Small Pricing Breakdown

At $0.50 per million input tokens and $1.50 per million output tokens, Magistral Small offers chain-of-thought reasoning at a fraction of the cost of premium reasoning models. The 3:1 output-to-input ratio is favorable for reasoning workloads, where models generate long chains of thinking. Here's how costs scale:

Monthly Volume	Input Cost	Output Cost	Total (1:3 input:output split)
1M tokens	$0.50	$1.50	$1.25
10M tokens	$5.00	$15.00	$12.50
100M tokens	$50.00	$150.00	$125.00
1B tokens	$500.00	$1,500.00	$1,250.00

At 100M tokens/month with a 1:3 input-to-output split (typical for reasoning tasks), Magistral Small costs $125. The same workload on Magistral Medium would cost $425, and on Claude Sonnet 5 it would cost $750+.

Magistral Small vs Other Budget Models

Model	Input $/M	Output $/M	Context	Reasoning	Provider
Mistral Small 4	$0.15	$0.60	128K	No	Mistral
GPT-5 nano	$0.05	$0.40	128K	No	OpenAI
DeepSeek V4 Pro	$0.14	$0.28	128K	Limited	DeepSeek
Gemini 3.1 Flash-Lite	$0.10	$0.40	1M	No	Google
Magistral Small	$0.50	$1.50	256K	Yes (CoT)	Mistral
Magistral Medium	$2.00	$5.00	256K	Yes (CoT)	Mistral
Claude Sonnet 5	$2.00	$10.00	200K	Yes	Anthropic

Magistral Small occupies a unique position: it's the cheapest model with dedicated chain-of-thought reasoning from a major provider. Non-reasoning budget models (Mistral Small 4, GPT-5 nano) are cheaper, but they lack the systematic thinking capability that makes reasoning models accurate on complex problems.

Magistral Reasoning Family: Small vs Medium

Mistral's Magistral line is their dedicated reasoning family — models specifically trained for chain-of-thought problem solving. Here's how the two tiers compare:

Feature	Magistral Small	Magistral Medium
Input price	$0.50/M	$2.00/M
Output price	$1.50/M	$5.00/M
Context window	256K	256K
Chain-of-thought	Yes	Yes
Reasoning quality	Good	Best
Cost multiplier	1x (baseline)	3.3x output, 4x input
Best for	Budget reasoning, high-volume analysis	Hard problems, production accuracy

Rule of thumb: Start with Magistral Small for reasoning tasks. Upgrade to Magistral Medium when accuracy on difficult problems is critical and you can justify 3-4x the cost. For non-reasoning tasks (chat, extraction, classification), use Mistral Small 4 ($0.15/$0.60) instead — it's 3x cheaper and handles those tasks well.

When to Use Magistral Small

Mathematical Problem Solving

Step-by-step math, algebra, statistics, and quantitative analysis. Chain-of-thought reasoning produces more accurate answers than direct-response models.

Logic & Analysis

Multi-step logical reasoning, deduction, and argument analysis. The model works through premises systematically before reaching conclusions.

Code Reasoning

Debugging, algorithm design, and code analysis that requires understanding execution flow. Not code generation — reasoning about what code does.

Structured Extraction

Extracting structured data from complex, unstructured sources. Reasoning chains help the model handle edge cases and ambiguous formats.

Complex Q&A

Questions requiring synthesis across multiple documents or domains. The 256K context lets you feed in extensive reference material.

Research Analysis

Evaluating evidence, comparing findings across papers, and drawing conclusions from large bodies of text. Budget-friendly alternative to premium reasoning models.

When NOT to Use Magistral Small

Magistral Small is optimized for reasoning, not general-purpose tasks. Consider alternatives when you need:

Simple chat or Q&A: Use Mistral Small 4 ($0.15/$0.60) or GPT-5 nano ($0.05/$0.40) — 3-10x cheaper for tasks that don't need reasoning
Content generation: Writing, marketing copy, creative content → use Claude Sonnet 5 ($2/$10) or Mistral Small 4 ($0.15/$0.60)
Highest reasoning accuracy: Hard math olympiad problems, complex code generation → use Magistral Medium ($2/$5) or Claude Opus 5 ($5/$25)
Massive context needs: 256K not enough → Gemini 3.1 Flash-Lite offers 1M context at $0.10/$0.40 (no reasoning)
Ultra-high volume, simple tasks: Classification, tagging, basic extraction → Qwen 3.7 Flash ($0.03/$0.13) is 10x cheaper

Real-World Cost: 25K Reasoning Tasks/Month

Consider a system that handles 25,000 reasoning tasks per month — math problems, logic analysis, or complex extraction. Each task averages 1,000 input tokens (problem statement + context) and 2,000 output tokens (reasoning chain + answer).

Model	Input Cost	Output Cost	Total/Month
Mistral Small 4 (no reasoning)	$3.75	$30.00	$33.75
Magistral Small	$12.50	$75.00	$87.50
Magistral Medium	$50.00	$250.00	$300.00
Claude Sonnet 5	$50.00	$500.00	$550.00
Claude Opus 5	$125.00	$1,250.00	$1,375.00

At $87.50/month for 25K reasoning tasks, Magistral Small costs 3.4x less than Magistral Medium and 6.3x less than Claude Sonnet 5 — while providing dedicated chain-of-thought reasoning that non-reasoning budget models lack.

Frequently Asked Questions

How much does Magistral Small cost?

$0.50 per million input tokens and $1.50 per million output tokens. The output-to-input ratio is 3:1 — lower than most models, which means output-heavy reasoning chains are relatively affordable compared to models like Claude Sonnet 5 ($2/$10) or GPT-5.4 ($2.50/$15).

What is the context window of Magistral Small?

256K tokens — double the 128K offered by Mistral Small 4 and most budget models. This makes it well-suited for complex reasoning tasks that require large context, such as multi-document analysis, long codebases, or detailed problem statements.

How does Magistral Small differ from Magistral Medium?

Magistral Medium ($2.00/$5.00) is 4x more expensive on input and 3.3x more expensive on output than Magistral Small ($0.50/$1.50). Both have chain-of-thought reasoning, but Medium produces higher-quality reasoning chains and handles more complex problems. Use Magistral Small for budget reasoning tasks; upgrade to Medium when accuracy on hard problems matters more than cost.

Is Magistral Small GDPR compliant?

Yes. Mistral is a French company headquartered in Paris with European API infrastructure. Magistral Small processes data on European data centers, making it a strong choice for EU-based companies that need GDPR-compliant AI reasoning without data leaving European jurisdiction.

What is Magistral Small good for?

Magistral Small excels at tasks that require step-by-step reasoning: mathematical problem-solving, logical analysis, code reasoning, structured data extraction, and complex Q&A. Its chain-of-thought capability makes it significantly better than non-reasoning budget models for tasks where accuracy depends on working through problems systematically.

Calculate Your Magistral Small Costs

Enter your token counts and see exactly what you'd pay. Compare against 93 other models instantly.

Open Calculator Compare Models

Mistral Small 4 Pricing — Mistral's general-purpose budget model at $0.15/$0.60
All Mistral Models — Full Mistral pricing and model comparison
GPT-5 nano Pricing — OpenAI's cheapest model at $0.05/$0.40
Gemini 3.1 Flash-Lite Pricing — Google's budget model at $0.10/$0.40
API Cost Calculator — Compare costs across 94 models
Model Leaderboard — Ranking by price, quality, and speed