Mistral AI 推出的 Devstral Medium 是一款专为长周期自主智能体 (Agent)、工具调用与自动化工作流打造的大模型，具备 128K 上下文窗口、75 Token/秒生成速度与 52/100 综合智能指数（每百万 Token $0.8 输入 / $2.4 输出）。
Technical Specifications
- Provider
- Mistral AI
- Category
- flagship
- Release Date
- 2025-07
- Batch API Discount
- 50% off async batch
- Consumer Subscription
- API Pay-as-you-go
- Reasoning Effort Control
- Configurable (Low / Med / High)
- Structured Outputs
- Yes
- Output Modalities
- text
- Official Documentation
- Visit Official Docs ↗
Data Privacy, Licensing & Deployment Governance
Official Provider Release Intelligence, API Quotas & Agentic Specs
- Official API Model ID (SDK String)
- devstral-medium-latest
- Supported SDK & API Protocol
- Mistral AI REST API (OpenAI-Compatible)
- Lifecycle & Deprecation Status
- Active GA (Guaranteed support through 2027+)
- Production Rate Limits (RPM & TPM)
- 5,000 – 15,000 RPM (Production Tier) · 800,000 – 2,000,000 TPM
- Long-Context Pricing Tier (>128K / >200K)
- Flat rate ($0.8/1M in) across full 128K window
- Prompt Cache Write & TTL Policy
- Cached input at $0.2/1M tokens
- Computer Use & Browser GUI Agent
- API / CLI Tool-Use & DOM Agent ready (no raw pixel desktop click loop)
- MCP & Parallel Tool Calling
- Native MCP Client/Server ready · Parallel function calling · 100% Strict JSON Schema
- Web Search Grounding & Code Sandbox
- Supported via external Tavily/Exa Search & E2B/Docker Sandbox tools
- Fine-Tuning & Distillation Support
- API Prompt Engineering & RAG Distillation
- Multimodal Tokenization Billing
- Text BPE (~4 chars/token)
- Weights Precision & Quantization
- Cloud-Hosted Proprietary (FP8 / BF16 Tensor Core Cluster)
- Minimum GPU VRAM Requirement
- Zero Local VRAM Required (100% Serverless Cloud API)
- Supported Inference Engines
- Managed Mistral AI Cloud API
- System Card & Safety Framework
- Mistral AI Production Safety & Alignment Verified
- Enterprise Compliance (SOC2 / HIPAA / ZDR)
- SOC2 Type II · GDPR · Zero Data Retention (ZDR) & HIPAA BAA Available on Enterprise Tier
Devstral Medium Cost, Context, Hardware & API Integration
For single-user chat, a $20/mo subscription caps your bill below this API volume.
Break-even point: ~152 requests/day equals $20/month.
Same Monthly Workload Cost Across Popular Reference Models
Live Prompt Paste Tokenizer & Cost Counter
Paste real system instructions, JSON schemas, or code to measure BPE token counts across tokenizers.
Hybrid Prompt Routing Cascade Simulator
Route routine queries to a fast lightweight model ($0.15/1M) and escalate only complex reasoning turns to Devstral Medium:
Benchmark Scores & 6-Axis Capability Radar
- SWE-bench Verified
- 51.5%
- GPQA Diamond
- 62%
- MMLU-Pro
- 80.5%
- HumanEval
- 92.4%
- Terminal-Bench
- 45%
- AIME
- 72%
- Long-Context Reasoning
- 78%
- Chatbot Arena Elo
- 1330pts
Strengths
- 提供 128K 上下文窗口与最高 33K 输出容量，核心基准表现优异（SWE-bench Verified 51.5% · GPQA Diamond 62%）
- 高达 75 Token/秒的生成吞吐量，首字延迟 (TTFT) 仅约 500 毫秒
- 极具竞争力的 API 定价（每百万 Token $0.8 输入 / $2.4 输出），原生支持提示词缓存、函数调用与结构化 JSON 输出
Limitations & Trade-offs
- 闭源专有云端模型，不支持下载权重进行离线本地物理隔离部署
- 在超长链路的前沿科研推导任务上，深度略逊于顶配重型旗舰推理模型
Editorial Assessment
Devstral Medium 凭借 52/100 的智能得分、128K 上下文容量与 75 Token/秒的响应速度，在 Mistral AI 模型矩阵中提供了极佳的性能与成本平衡。
Best for: 专为长周期自主智能体 (Agent)、工具调用与自动化工作流打造的任务、企业级 RAG 检索增强生成、自动化智能体工作流与高可靠 API 集成。
Compare Devstral Medium
Frequently Asked Questions about Devstral Medium
How much does the Devstral Medium API cost per 1M tokens?+
Devstral Medium by Mistral AI is priced at $0.8 per 1M input tokens and $2.4 per 1M output tokens. With prompt caching enabled, cached input tokens cost $0.2 per 1M tokens, and asynchronous Batch API requests receive a 50% discount.
Should I use a $20/month subscription or the Devstral Medium pay-as-you-go API?+
A flat $20/month consumer subscription is predictable for daily interactive chat and manual coding. However, if you are building software, running automated workflows, or sending low-to-moderate volume with prompt caching enabled, the pay-as-you-go API is typically far more cost-effective and offers higher rate limits without manual UI caps.
How many words and pages fit in Devstral Medium's context window?+
Devstral Medium features a 128K tokens context window, which equals approximately 96,000 English words or ~256 standard single-spaced pages in a single prompt. Its maximum output generation per turn is 32,768 tokens.
Can I run Devstral Medium locally on my own GPU (Open Weights)?+
No, Devstral Medium is a proprietary closed-weights model hosted by Mistral AI. It cannot be downloaded for local GPU execution and must be accessed via Mistral AI's official API or cloud platform.
Does Mistral AI train on data sent to the Devstral Medium API, and is commercial use allowed?+
API payloads are excluded from model training by default. License tier: Proprietary Commercial API (Mistral AI Terms). Full commercial ownership of generated outputs. Supported deployment channels include: Mistral AI Cloud API & Enterprise Endpoints.
How does Devstral Medium perform on coding and reasoning benchmarks?+
Devstral Medium achieves a composite Intelligence Index of 52/100, scoring 51.5% on SWE-bench Verified, 62% on GPQA Diamond, 80.5% on MMLU-Pro, and 1330 pts on Chatbot Arena Elo.
What is Devstral Medium best used for?+
Devstral Medium is best suited for: 专为长周期自主智能体 (Agent)、工具调用与自动化工作流打造的任务、企业级 RAG 检索增强生成、自动化智能体工作流与高可靠 API 集成。. Assessment: Devstral Medium 凭借 52/100 的智能得分、128K 上下文容量与 75 Token/秒的响应速度，在 Mistral AI 模型矩阵中提供了极佳的性能与成本平衡。
Mistral AI Family Lineage, Routing Tiers & Upgrade Path
Step down to a lighter sibling for high-volume routing or step up to Mistral AI's flagship reasoning tier:
Mistral Medium 3.5
Mistral Small 4
Devstral 2
Devstral Small
Mistral Small 3.2
Mistral Medium 3
AI Glossary & Architecture Concepts
View full glossary →Key machine learning concepts and benchmarks relevant to Devstral Medium:
The basic unit of text that a language model processes, typically representing about 3 to 4 English characters or roughly 0.75 words. For example, the sentence "Artificial intelligence is fast" breaks down into about 5 tokens.
A standardized evaluation suite used to objectively measure and rank an AI model's capabilities across specific tasks — such as SWE-bench Verified for software engineering, AIME for competitive math, or MMLU for general knowledge.
The maximum number of tokens a model can consider at once — both the input it reads and the output it writes. A 1,000,000-token context window can hold roughly 750,000 English words in a single request.
An AI model trained via reinforcement learning to spend dynamic compute thinking through complex logic, math, and architectural problems before outputting a solution. Examples include OpenAI o1, o3, and DeepSeek R1.
A step-by-step reasoning technique where an AI model generates intermediate problem-solving thoughts before arriving at a final answer. Native reasoning models like OpenAI o3 and DeepSeek R1 use CoT to solve complex math and coding challenges.
The ability of a language model to recognize when an external capability is needed and output structured JSON arguments to invoke external APIs, databases, or terminal commands. The runtime executes the tool and passes results back to the model.
Benchmark scores, token prices, and rate limits evolve rapidly as AI labs release model updates. Always verify current specifications and pricing on the official provider documentation.