Context Window
131,072
Tokens
Model Cost Profile
Developer: mistralai
Pricing updated Mar 11, 2026
Live Pricing
Input: $0.4000
Output: $2.00
Pricing via OpenRouter API ยท Last synced Mar 11, 2026 ยท MMLU score via public benchmark data
Mistral Medium 3, developed by mistralai, offers an expansive context window of 131,072 tokens, making it ideal for applications requiring in-depth analysis or long-form content generation. With an input price of $0.40 per 1 million tokens and an output price of $2.00 per million tokens, teams can effectively manage their budget while leveraging the model for complex tasks such as document summarization or interactive dialogue systems. This model is particularly suited for enterprises needing to process large datasets or maintain context over extended conversations, ensuring high-quality outputs with cost efficiency.
Context Window
131,072
Tokens
Input Price / 1M
$0.4000
Prompt tokens
Output Price / 1M
$2.00
Completion tokens
Intelligence (MMLU)
68.3
Massive Multitask Language Understanding
Standardized evaluation scores for Mistral: Mistral Medium 3.
| Benchmark | Score | Rank | Source |
|---|---|---|---|
| MMLU | 68.3 | #93 of 121 | artificial_analysis |
| Usage Type | Price / 1M Tokens |
|---|---|
| Input (Prompt) | $0.4000 |
| Output (Completion) | $2.00 |
Price History
Current Input / 1M
$0.4000
Current Output / 1M
$2.00
Estimate monthly spend for Mistral: Mistral Medium 3 based on your workload.
Estimated Monthly Cost
$34
25M input + 12M output tokens
Quick links for cost-down decisions before production rollout.
Common pricing and benchmark questions for Mistral: Mistral Medium 3.
Mistral: Mistral Medium 3 input pricing is $0.4000 per 1M tokens based on the latest synced provider data.
Mistral: Mistral Medium 3 output pricing is $2.00 per 1M tokens based on the latest synced provider data.
Mistral: Mistral Medium 3 supports a context window of 131,072 tokens.
Use the comparison links on this page to open direct model-vs-model pricing and benchmark pages, then evaluate monthly spend projections for your workload.