Devstral Small vs Codestral
Benchmarks, API Token Pricing, Context Window & Speed Comparison
Related topics (12)
Mistral AI 的 Devstral Small 与 Mistral AI 的 Codestral 是当前备受关注的两款主力 AI 模型。Devstral Small 提供 128K Token 上下文窗口，API 定价为每百万 Token 输入 $0.15 / 输出 $0.45；而 Codestral 提供 256K Token 上下文窗口，定价为每百万 Token 输入 $0.20 / 输出 $0.60。以下为两者的基准测试、延迟与成本全面对比。
切换并对比任意模型
Select any AI models from our verified benchmark database to generate live side-by-side matrices.
Devstral Small
自主智能体与工具专家Mistral AI 推出的 Devstral Small 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、95 Token/秒生成速度与 45/100 综合智能指数（每百万 Token $0.15 输入 / $0.45 输出）。
$0.15 / 1M
$0.45 / 1M
128K tokens
42%
Codestral
开源权重编程专家Mistral AI 推出的 Codestral 是一款面向本地私有化部署与代码生成的开源权重编程大模型，具备 256K 上下文窗口、110 Token/秒生成速度与 40/100 综合智能指数（每百万 Token $0.2 输入 / $0.6 输出）。
$0.20 / 1M
$0.60 / 1M
256K tokens
28%
核心差异一览
- Devstral Small is 1.3x cheaper for API usage (input + output tokens combined).
- Codestral has a larger context window (256K vs 128K tokens).
- Codestral is faster at 110 tok/s vs 95 tok/s.
- Devstral Small leads on SWE-bench coding (42.0% vs 28.0%).
| Feature | Devstral Small | Codestral |
|---|---|---|
Intelligence Index Composite reasoning & knowledge score (0-100) | 45 | 40 |
Provider | Mistral AI | Mistral AI |
Input Price Cost per million input tokens | $0.15 / 1M | $0.20 / 1M |
Output Price Cost per million output tokens | $0.45 / 1M | $0.60 / 1M |
Blended Price Weighted 7:2:1 cache:input:output per 1M | $0.11 / 1M | $0.14 / 1M |
Context Window Maximum input + output tokens | 128K tokens~192 A4 pages of text | 256K tokens~384 A4 pages of text |
Max Output Tokens Per single request | 16K | 8K |
Throughput Speed Tokens generated per second | 95 tok/s | 110 tok/s |
SWE-bench Verified Real-world software engineering | 42% | 28% |
GPQA Diamond Graduate-level reasoning | 50% | 42% |
Cached Input Price Prompt caching discount per 1M tokens | $0.05 / 1M | $0.05 / 1M |
各项指标谁更胜一筹？
At-a-glance decision framework权威基准测试得分对比
Official evaluation metrics verified by independent third-party research labs.
Industry benchmark evaluating real-world GitHub issue resolution and multi-file debugging.
Challenging questions created by domain experts to test complex PhD-level reasoning.
Hardened reasoning and academic knowledge benchmark with reduced lucky-guess probability.
Crowdsourced human preference evaluation based on blind, head-to-head model battles.
Autonomous multi-step tasks executed inside a real command-line terminal environment.
American Invitational Mathematics Examination — multi-step Olympiad-level reasoning.
Accuracy when reasoning over very large inputs that fill the model context window.
Tokens emitted per second during active generation streaming.
延迟与响应速度
Time to first token (TTFT) and estimated end-to-end response time for a typical 500-token response.
实时 Token 成本模拟器
Adjust sliders to calculate projected monthly cloud API spending with prompt caching discounts.
Devstral Small · Strengths & Key Advantages
- 提供 128K 上下文窗口与最高 16K 输出容量，核心基准表现优异（SWE-bench Verified 42% · GPQA Diamond 50%）
- 高达 95 Token/秒的生成吞吐量，首字延迟 (TTFT) 仅约 420 毫秒
- 开放模型权重 (24B)，支持通过 vLLM、SGLang 或 Ollama 在私有 GPU 集群上本地部署与微调
Considerations & Trade-offs
- 在本地以满血精度运行需要较高的 GPU 显存 (VRAM) 与运维基础设施
- 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
Codestral · Strengths & Key Advantages
- 提供 256K 上下文窗口与最高 8K 输出容量，核心基准表现优异（SWE-bench Verified 28% · GPQA Diamond 42%）
- 高达 110 Token/秒的生成吞吐量，首字延迟 (TTFT) 仅约 350 毫秒
- 开放模型权重 (22B)，支持通过 vLLM、SGLang 或 Ollama 在私有 GPU 集群上本地部署与微调
Considerations & Trade-offs
- 在本地以满血精度运行需要较高的 GPU 显存 (VRAM) 与运维基础设施
- 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
Token Use & Task Economics
Estimated cost and token consumption to complete one representative agentic task.
| Metric | Devstral Small | Codestral |
|---|---|---|
Cost per Task Estimated USD to finish one task | $0.14 | $0.20 |
Output Tokens / Task Tokens generated per task | 30K | 24K |
Reasoning Tokens / Task Thinking tokens burned per task | 0K | 0K |
技术规格参数
Architecture, release timeline, capabilities, and modalities.
| Feature | Devstral Small | Codestral |
|---|---|---|
| Official API Model ID (SDK) | devstral-small-latest | codestral-latest |
| API Protocol | OpenAI-Compatible REST / vLLM / SGLang / HuggingFace TGI | OpenAI-Compatible REST / vLLM / SGLang / HuggingFace TGI |
| Release Date | 2025-07 | 2024-05 |
| Knowledge Cutoff | May 2025 | May 2024 |
| Architecture | Code-Optimized Autoregressive Transformer | Code-Optimized Transformer with 32K Window |
| Parameters | 24B | 22B |
| Enterprise Rate Limits (RPM / TPM) | Unlimited (Hardware-bound on self-hosted GPU) / 10,000 RPM Cloud · Unlimited local throughput / 10M+ TPM Managed Cloud | Unlimited (Hardware-bound on self-hosted GPU) / 10,000 RPM Cloud · Unlimited local throughput / 10M+ TPM Managed Cloud |
| Long-Context Tier (>128K/200K) | Flat rate ($0.15/1M in) across full 128K window | Flat rate ($0.2/1M in) across full 256K window |
| Prompt Cache Policy & TTL | Cached input at $0.05/1M tokens | Cached input at $0.05/1M tokens |
| Computer Use / GUI Agent | API / CLI Tool-Use & DOM Agent ready (no raw pixel desktop click loop) | 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 | Native MCP Client/Server ready · Parallel function calling · 100% Strict JSON Schema |
| Web Search Grounding & Sandbox | Supported via external Tavily/Exa Search & E2B/Docker Sandbox tools | Supported via external Tavily/Exa Search & E2B/Docker Sandbox tools |
| Official Fine-Tuning Support | Full Weights SFT, LoRA, QLoRA, DPO & GRPO RL (Unsloth / Axolotl / Llama-Factory) | Full Weights SFT, LoRA, QLoRA, DPO & GRPO RL (Unsloth / Axolotl / Llama-Factory) |
| Multimodal Token Billing | Text BPE (~4 chars/token) | Text BPE (~4 chars/token) |
| Quantization & Min GPU VRAM | BF16 / FP16, FP8 Dynamic, INT4 AWQ / GPTQ, GGUF (Q4_K_M / Q8_0) · 15 GB VRAM (4-bit Quantized) · 52 GB VRAM (Native BF16) | BF16 / FP16, FP8 Dynamic, INT4 AWQ / GPTQ, GGUF (Q4_K_M / Q8_0) · 14 GB VRAM (4-bit Quantized) · 47 GB VRAM (Native BF16) |
| Safety & Frontier Classification | Open-Weights System Card + Customizable Guardrail Layer (Llama Guard / ShieldGemma) | Open-Weights System Card + Customizable Guardrail Layer (Llama Guard / ShieldGemma) |
| Enterprise Compliance (ZDR/HIPAA/SOC2) | 100% Air-Gapped On-Premises / Private VPC · Full HIPAA, GDPR & SOC2 Control | 100% Air-Gapped On-Premises / Private VPC · Full HIPAA, GDPR & SOC2 Control |
| Input Modalities | text | text |
| Output Modalities | text | text |
Devstral Small vs Codestral: 常见问题解答 (FAQ)
Which is more intelligent, Devstral Small or Codestral?
Devstral Small is more intelligent. It scores 45 compared with 40 on the composite Intelligence Index.
Which model is faster?
Codestral is faster, generating 110 tokens per second compared with 95 tokens per second.
Which model is cheaper to run?
Devstral Small is cheaper at $0.11 per 1M tokens blended, compared with $0.14 (7:2:1 cache:input:output ratio).
Which model has lower latency?
Codestral has lower latency, with a time to first token of 0.35s compared with 0.42s.
Which model has a larger context window?
Codestral has a larger context window, supporting 256K tokens compared with 128K tokens.
综合结论
综合来看，Devstral Small（适合：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。）与 Codestral（适合：面向本地私有化部署与代码生成的开源权重编程任务、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。）各具优势：Devstral Small 在综合智能与编程基准测试中表现更强，Codestral 拥有更快的 Token 生成速度，而 Devstral Small 则在规模化 API 部署中具备更高的成本效益。
常见问题 (FAQ)
在编程和深度推理方面，Devstral Small 和 Codestral 哪个更强？+
在综合智能指数与编程基准评测中，Devstral Small 表现占优。建议结合上方的 SWE-bench Verified 与 GPQA Diamond 跑分根据具体场景选择。
Devstral Small 与 Codestral 的 API 价格哪个更便宜？+
Devstral Small 的 API 价格为输入 $0.15 / 输出 $0.45（每百万 Token），Codestral 为输入 $0.20 / 输出 $0.60。综合混合成本来看，Devstral Small 更具性价比。
哪款模型拥有更大的上下文窗口和更快的响应速度？+
Devstral Small 支持 128K Token 上下文窗口，而 Codestral 支持 256K Token。在流式生成速度方面，Codestral 的吞吐表现更快。
Understand key metrics like context window, token pricing, reasoning mode, and SWE-bench scores in our comprehensive AI glossary.
Related reading
- Codestral vs DeepSeek-Coder-V2 236B：基准测试、API 价格与速度全面对比 (2026)对比
- Devstral Small vs GPT-5.1 Codex mini：基准测试、API 价格与速度全面对比 (2026)对比
- Aion-3.0 vs Aion-Mini：基准测试、API 价格与速度全面对比 (2026)对比
- Aion-3.0 vs Gemini 3.6 Flash：基准测试、API 价格与速度全面对比 (2026)对比
- Aion-Mini vs Laguna XS 2.1：基准测试、API 价格与速度全面对比 (2026)对比
- Aya Expanse 32B vs Llama 3.2 90B Vision：基准测试、API 价格与速度全面对比 (2026)对比
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.
数据来源: Mistral AI 与 Mistral AI 官方 API 文档、模型发布说明及独立基准测试报告 (2026)。