GetInfoUs
Head-to-Head Model Comparison

Mistral Small 3.2 vs Mistral Medium 3

Benchmarks, API Token Pricing, Context Window & Speed Comparison

由 GetInfoUs 研究团队审核最后审核日期：October 2026编辑政策
分享
Related topics (12)
AI CompareAI BenchmarkToken PricingModel CostModel SpecsSWE-bench CodingModel VersusWhich AI ModelLLM Speed & LatencyReasoning ModelsExtended ThinkingPrompt Caching ROI
Mistral Small 3.2 vs Mistral Medium 3: Benchmarks, Pricing & Speed Comparison

Mistral AI 的 Mistral Small 3.2 与 Mistral AI 的 Mistral Medium 3 是当前备受关注的两款主力 AI 模型。Mistral Small 3.2 提供 128K Token 上下文窗口，API 定价为每百万 Token 输入 $0.10 / 输出 $0.30；而 Mistral Medium 3 提供 131.1K Token 上下文窗口，定价为每百万 Token 输入 $0.40 / 输出 $2.00。以下为两者的基准测试、延迟与成本全面对比。

模型直接对比 (VS)

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.

Profile
Profile
Multi-Model Arena Stack:
5-Model Matrix Workbench
Popular Matchups:
Mistral AI

Mistral Small 3.2

开源权重基础大模型

Mistral AI 推出的 Mistral Small 3.2 是一款支持私有云服务器自托管、微调与完全数据主权的开源权重基础大模型，具备 128K 上下文窗口、100 Token/秒生成速度与 43/100 综合智能指数（每百万 Token $0.1 输入 / $0.3 输出）。

Intelligence Index43 / 100
Input Token:

$0.10 / 1M

Output Token:

$0.30 / 1M

Context Window:

128K tokens

SWE-bench Verified:

34%

Mistral AI

Mistral Medium 3

开源权重基础大模型

Mistral AI 推出的 Mistral Medium 3 是一款支持私有云服务器自托管、微调与完全数据主权的开源权重基础大模型，具备 131K 上下文窗口、85 Token/秒生成速度与 47/100 综合智能指数（每百万 Token $0.4 输入 / $2 输出）。

Intelligence Index47 / 100
Input Token:

$0.40 / 1M

Output Token:

$2.00 / 1M

Context Window:

131.1K tokens

SWE-bench Verified:

40.5%

核心差异一览

  • Mistral Small 3.2 is 5.2x cheaper for API usage (input + output tokens combined).
  • Mistral Medium 3 has a larger context window (131K vs 128K tokens).
  • Mistral Small 3.2 is faster at 100 tok/s vs 85 tok/s.
  • Mistral Medium 3 leads on SWE-bench coding (40.5% vs 34.0%).
Need definitions for token pricing, context, or benchmarks?Explore AI Glossary →
FeatureMistral Small 3.2Mistral Medium 3
Intelligence Index
Composite reasoning & knowledge score (0-100)
4347
Provider
Mistral AIMistral AI
Input Price
Cost per million input tokens
$0.10 / 1M$0.40 / 1M
Output Price
Cost per million output tokens
$0.30 / 1M$2.00 / 1M
Blended Price
Weighted 7:2:1 cache:input:output per 1M
$0.09 / 1M$0.35 / 1M
Context Window
Maximum input + output tokens
128K tokens~192 A4 pages of text131.1K tokens~197 A4 pages of text
Max Output Tokens
Per single request
8K16K
Throughput Speed
Tokens generated per second
100 tok/s85 tok/s
SWE-bench Verified
Real-world software engineering
34%40.5%
GPQA Diamond
Graduate-level reasoning
48.5%57.5%
Cached Input Price
Prompt caching discount per 1M tokens
$0.05 / 1M$0.10 / 1M

各项指标谁更胜一筹？

At-a-glance decision framework
Coding
Mistral Medium 3
+6.5% SWE-bench lead
Cost
Mistral Small 3.2
5.2x cheaper monthly runtime
Speed
Mistral Small 3.2
+18% faster generation
Context
Mistral Medium 3
131.1K tokens window

权威基准测试得分对比

Official evaluation metrics verified by independent third-party research labs.

Mistral Small 3.2Mistral Medium 3
SWE-bench Verified (Autonomous Coding)

Industry benchmark evaluating real-world GitHub issue resolution and multi-file debugging.

34%vs40.5%
GPQA Diamond (Doctorate-Level Science & Math)

Challenging questions created by domain experts to test complex PhD-level reasoning.

48.5%vs57.5%
MMLU-Pro (Multi-task Language Understanding)

Hardened reasoning and academic knowledge benchmark with reduced lucky-guess probability.

72.5%vs79.5%
LMSYS Chatbot Arena Elo

Crowdsourced human preference evaluation based on blind, head-to-head model battles.

1255ptsvs1310pts
Terminal-Bench (Agentic Coding & Terminal Use)

Autonomous multi-step tasks executed inside a real command-line terminal environment.

29%vs36%
AIME (Advanced Math Competition)

American Invitational Mathematics Examination — multi-step Olympiad-level reasoning.

55%vs68%
Long-Context Reasoning

Accuracy when reasoning over very large inputs that fill the model context window.

68%vs77%
Generation Speed (Throughput)

Tokens emitted per second during active generation streaming.

100 tok/s vs85 tok/s

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.

Time to First Token (TTFT)Lower is better
Mistral Small 3.20.40s
Mistral Medium 30.48s
Est. End-to-End Response (500 tokens)TTFT + generation time
Mistral Small 3.25.40s
Mistral Medium 36.36s

实时 Token 成本模拟器

Adjust sliders to calculate projected monthly cloud API spending with prompt caching discounts.

100K / day
10K500K1M
30K / day
5K150K300K
50% cached
0% (Cold)50%90% (Warm)
Mistral Small 3.2 Monthly API Bill
$0.50 / month
$0.1/1M in · $0.3/1M out
Mistral Medium 3 Monthly API Bill
$2.55 / month
$0.4/1M in · $2/1M out
Mistral Small 3.2 saves approximately $2.06 / month (81%) under this workload.5.2x Savings

Mistral Small 3.2 · Strengths & Key Advantages

  • 提供 128K 上下文窗口与最高 8K 输出容量，核心基准表现优异（SWE-bench Verified 34% · GPQA Diamond 48.5%）
  • 高达 100 Token/秒的生成吞吐量，首字延迟 (TTFT) 仅约 400 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

Mistral Medium 3 · Strengths & Key Advantages

  • 提供 131K 上下文窗口与最高 16K 输出容量，核心基准表现优异（SWE-bench Verified 40.5% · GPQA Diamond 57.5%）
  • 高达 85 Token/秒的生成吞吐量，首字延迟 (TTFT) 仅约 480 毫秒
  • 开放模型权重 (Confidential (~70B))，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.

MetricMistral Small 3.2Mistral Medium 3
Cost per Task
Estimated USD to finish one task
$0.09$0.35
Output Tokens / Task
Tokens generated per task
25K35K
Reasoning Tokens / Task
Thinking tokens burned per task
0K0K

技术规格参数

Architecture, release timeline, capabilities, and modalities.

FeatureMistral Small 3.2Mistral Medium 3
Official API Model ID (SDK)mistral-small-3-2-latestmistral-medium-3-latest
API ProtocolOpenAI-Compatible REST / vLLM / SGLang / HuggingFace TGIOpenAI-Compatible REST / vLLM / SGLang / HuggingFace TGI
Release Date2025-062025-05
Knowledge CutoffApril 2025March 2025
ArchitectureCompact Vision-Language TransformerEnterprise Multimodal Transformer
Parameters24BConfidential (~70B)
Enterprise Rate Limits (RPM / TPM)Unlimited (Hardware-bound on self-hosted GPU) / 10,000 RPM Cloud · Unlimited local throughput / 10M+ TPM Managed CloudUnlimited (Hardware-bound on self-hosted GPU) / 10,000 RPM Cloud · Unlimited local throughput / 10M+ TPM Managed Cloud
Long-Context Tier (>128K/200K)Flat rate ($0.1/1M in) across full 128K windowFlat rate ($0.4/1M in) across full 131K window
Prompt Cache Policy & TTLCached input at $0.05/1M tokensCached input at $0.1/1M tokens
Computer Use / GUI AgentAPI / 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 CallingNative MCP Client/Server ready · Parallel function calling · 100% Strict JSON SchemaNative MCP Client/Server ready · Parallel function calling · 100% Strict JSON Schema
Web Search Grounding & SandboxSupported via external Tavily/Exa Search & E2B/Docker Sandbox toolsSupported via external Tavily/Exa Search & E2B/Docker Sandbox tools
Official Fine-Tuning SupportFull 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 BillingText BPE (~4 chars/token) · Vision (~258–765 tokens per image tile)Text BPE (~4 chars/token) · Vision (~258–765 tokens per image tile)
Quantization & Min GPU VRAMBF16 / 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) · 43 GB VRAM (4-bit Quantized) · 151 GB VRAM (Native BF16)
Safety & Frontier ClassificationOpen-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 Control100% Air-Gapped On-Premises / Private VPC · Full HIPAA, GDPR & SOC2 Control
Input Modalitiestext, visiontext, vision
Output Modalitiestexttext

Mistral Small 3.2 vs Mistral Medium 3: 常见问题解答 (FAQ)

Which is more intelligent, Mistral Small 3.2 or Mistral Medium 3?

Mistral Medium 3 is more intelligent. It scores 47 compared with 43 on the composite Intelligence Index.

Which model is faster?

Mistral Small 3.2 is faster, generating 100 tokens per second compared with 85 tokens per second.

Which model is cheaper to run?

Mistral Small 3.2 is cheaper at $0.09 per 1M tokens blended, compared with $0.35 (7:2:1 cache:input:output ratio).

Which model has lower latency?

Mistral Small 3.2 has lower latency, with a time to first token of 0.40s compared with 0.48s.

Which model has a larger context window?

Mistral Medium 3 has a larger context window, supporting 131.1K tokens compared with 128K tokens.

综合结论

综合来看，Mistral Small 3.2（适合：支持私有云服务器自托管、微调与完全数据主权的开源权重基础任务、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 Mistral Medium 3（适合：支持私有云服务器自托管、微调与完全数据主权的开源权重基础任务、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：Mistral Medium 3 在综合智能与编程基准测试中表现更强，Mistral Small 3.2 拥有更快的 Token 生成速度，而 Mistral Small 3.2 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，Mistral Small 3.2 和 Mistral Medium 3 哪个更强？+

在综合智能指数与编程基准评测中，Mistral Medium 3 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。

Mistral Small 3.2 与 Mistral Medium 3 的 API 价格哪个更便宜？+

Mistral Small 3.2 的 API 价格为输入 $0.10 / 输出 $0.30（每百万 Token），Mistral Medium 3 为输入 $0.40 / 输出 $2.00。综合混合成本来看，Mistral Small 3.2 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？+

Mistral Small 3.2 支持 128K Token 上下文窗口，而 Mistral Medium 3 支持 131.1K Token。在流式生成速度方面，Mistral Small 3.2 的吞吐表现更快。

AI Terminology & Guide
Explore Glossary →

Understand key metrics like context window, token pricing, reasoning mode, and SWE-bench scores in our comprehensive AI glossary.

Tokens & PricingContext WindowAI BenchmarksReasoning ModelsTool CallingModel Context Protocol (MCP)

Related reading

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)。