Mistral Medium 3.5 API Pricing: Agentic Workhorse at $1.50/$7.50

Mistral Medium 3.5 is designed for long-horizon agentic tasks, synchronous tool-calling, and agentic coding. At $1.50/$7.50, it competes directly with Claude Sonnet 5 for agent workloads.

Updated Aug 21, 2026 · 95 models tracked across 11 providers

TL;DR

🇪🇺 EU Data Sovereignty: Mistral is headquartered in Paris with European data centers. Medium 3.5 is ideal for EU-based companies needing GDPR-compliant AI for agentic workflows. A 10% surcharge applies for EU regional inference endpoints.

Mistral Medium 3.5 Pricing Breakdown

Monthly costs at different token volumes (50/50 input/output split):

Monthly Volume Input Cost Output Cost Total (50/50)
1M tokens$1.50$7.50$4.50
10M tokens$15.00$75.00$45.00
100M tokens$150.00$750.00$450.00
1B tokens$1,500.00$7,500.00$4,500.00

At 100M tokens/month with a 50/50 split, Mistral Medium 3.5 costs $450. Cached input tokens get a 90% discount ($0.15/M), which can significantly reduce costs for repetitive agentic workflows.

Mistral Model Family: Medium 3.5 in Context

Model Input $/M Output $/M Context Best For
Mistral Small 4 $0.15 $0.60 128K Budget tasks, high throughput
Mistral Large 3 $0.50 $1.50 262K General tasks, long context
Mistral Medium 3.5 $1.50 $7.50 128K Agentic coding, tool-calling
Codestral $0.30 $0.90 256K Code completion, fill-in-middle

Decision guide: Choose Small 4 ($0.15/$0.60) for budget tasks. Choose Large 3 ($0.50/$1.50) when context length (262K) matters most. Choose Medium 3.5 ($1.50/$7.50) for agentic workflows requiring tool-calling and iterative reasoning. Choose Codestral ($0.30/$0.90) for code-specific completion tasks.

Mistral Medium 3.5 vs Other Mid-Tier Models

Model Input $/M Output $/M Context Provider
GPT-5.4 $2.50 $15.00 400K OpenAI
Mistral Medium 3.5 $1.50 $7.50 128K Mistral
Claude Sonnet 5 $2.00 $10.00 1M Anthropic
Gemini 3.1 Pro $2.00 $12.00 1M Google
Grok 4.5 $2.00 $6.00 500K xAI
DeepSeek V4 Pro $0.66 $1.98 1M DeepSeek

Mistral Medium 3.5 is the cheapest mid-tier model at $1.50/$7.50 — 25% cheaper than Sonnet 5 on both input and output. Grok 4.5 ($2/$6) has cheaper output but more expensive input. DeepSeek V4 Pro ($0.66/$1.98) is far cheaper but may not match Medium 3.5's agentic capabilities.

When to Use Mistral Medium 3.5

🤖 Agentic Coding

Build coding agents that write, test, and iterate on code. Medium 3.5's synchronous tool-calling makes it ideal for multi-step development workflows.

🔧 Tool-Calling Workflows

Automate tasks that require calling APIs, databases, or external services. Designed for synchronous tool-calling with reliable output formatting.

📊 Enterprise Automation

Build internal tools that process data, generate reports, and make decisions. EU data sovereignty makes it ideal for European enterprises.

💬 Multi-Turn Conversations

Build chatbots that maintain context across long conversations. 128K context handles extended dialogues with ease.

📝 Document Processing

Analyze, summarize, and extract information from documents. Balanced pricing makes it cost-effective for mixed input/output workloads.

🔄 Workflow Orchestration

Coordinate multiple AI agents or services. Medium 3.5's reasoning capabilities make it a strong orchestrator for complex multi-step tasks.

Real-World Cost: 10K Agentic Coding Tasks/Month

Suppose you're building an AI coding assistant that handles 10,000 coding tasks per month, averaging 2,000 input tokens (code context + instructions) and 1,500 output tokens (code changes + explanations) per request:

ModelInput CostOutput CostMonthly Total
DeepSeek V4 Pro$13.20$29.70$42.90
Mistral Large 3$10.00$22.50$32.50
Mistral Medium 3.5$30.00$112.50$142.50
Claude Sonnet 5$40.00$150.00$190.00
GPT-5.4$50.00$225.00$275.00

Mistral Medium 3.5 at $142.50/month is 25% cheaper than Claude Sonnet 5 ($190) and 48% cheaper than GPT-5.4 ($275) for agentic coding tasks. If quality is sufficient, DeepSeek V4 Pro ($42.90) or Mistral Large 3 ($32.50) offer even lower costs.

Compare 95 AI Models Side by Side

Mistral Medium 3.5 is one of 95 models tracked on APIpulse. Compare pricing, context windows, and features across 11 providers.

Frequently Asked Questions

How much does Mistral Medium 3.5 cost?
Mistral Medium 3.5 costs $1.50 per million input tokens and $7.50 per million output tokens. The output-to-input ratio is 5:1 — typical for mid-tier models. At 100M tokens/month with a 50/50 split, that's $450/month.
What is the context window of Mistral Medium 3.5?
Mistral Medium 3.5 has a 128K token context window. While smaller than Mistral Large 3 (262K) or Claude Sonnet 5 (1M), 128K is sufficient for most agentic workflows, multi-turn conversations, and code generation tasks.
How does Mistral Medium 3.5 compare to Claude Sonnet 5?
Mistral Medium 3.5 ($1.50/$7.50) costs 25% less on input and 25% less on output than Claude Sonnet 5 ($2.00/$10.00). Both are designed for agentic coding and tool-calling. Sonnet 5 has a larger context window (1M vs 128K) and may outperform on complex tasks, but Medium 3.5 offers competitive pricing with EU data sovereignty.
What is Mistral Medium 3.5 good for?
Mistral Medium 3.5 is designed for long-horizon agentic tasks, synchronous tool-calling, and agentic coding. It excels at multi-step workflows that require planning, tool use, and iterative reasoning. It's also a good choice for enterprise applications requiring EU data sovereignty.
Is Mistral Medium 3.5 GDPR compliant?
Yes. Mistral is a French company headquartered in Paris, and their API infrastructure runs on European data centers. Mistral Medium 3.5 is a strong choice for EU-based companies that need GDPR-compliant AI for agentic workflows. A 10% surcharge applies for EU regional inference endpoints.
How does Mistral Medium 3.5 compare to Mistral Large 3?
Mistral Medium 3.5 ($1.50/$7.50, 128K) costs 3x more on input and 5x more on output than Mistral Large 3 ($0.50/$1.50, 262K). However, Medium 3.5 is designed for agentic tasks with synchronous tool-calling, while Large 3 is a general-purpose model. Choose Medium 3.5 for agentic workflows; choose Large 3 for general tasks or when context length matters more.

Related Pages