Mistral Large 4 is a frontier multimodal (text and image input) model from Mistral AI built for reasoning, coding, and agentic workloads. It offers a 1M-token context window with up to 256K output tokens, and supports tool calling and structured outputs.
提供商
不同公司托管同一模型。OpenRouter 根据您选择的路由模式将请求路由到其中一家公司:Balanced(价格 + 速度)、Nitro(速度最快)或 Floor(价格最低)。
| US$0.70 | US$2.10 | US$0.070 | 0.85s | 55 tok/s | ||
| US$0.77 | US$2.31 | US$0.077 | 0.80s | 50 tok/s |

- 输入 /百万
- US$0.70
- 输出 /百万
- US$2.10
- 缓存读取 /百万
- US$0.07
- 延迟
- 0.85s
- 吞吐量
- 55 tok/s
- 正常运行时间
- 99.61%

- 输入 /百万
- US$0.77
- 输出 /百万
- US$2.31
- 缓存读取 /百万
- US$0.077
- 延迟
- 0.80s
- 吞吐量
- 50 tok/s
- 正常运行时间
- 99.29%
定价
客户为此模型实际支付的平均价格,与提供商公布的价格并列展示。由于缓存和折扣,实际支付的价格往往远低于标价。
加权平均值
Mistral Large 4 — 价格历史
| US$0.132 | US$2.10 | 90.1% | 90.4% | ||
| US$0.225 | US$2.31 | 78.7% | 9.6% |
- 有效输入价格 US$0.132 / 实际输出 US$2.10token 占比90.4%
- 有效输入价格 US$0.225 / 实际输出 US$2.31token 占比9.6%
性能
吞吐量是模型生成内容的速度(每秒 token 数,越高越好)。延迟是完整的往返时间(越低越好)。TTFT 是首个 token 的响应时间,即您看到内容出现前需要等待的时间(越低越好)。
P50,所有提供商中最佳
P50,最佳提供商
正常运行时间
正常运行时间是指过去 3 天中至少有一家提供商响应请求的时间百分比。可用性是指成功提供推理服务的时间百分比。OpenRouter 会持续监控,并在某家提供商返回错误时使用次优提供商。
过去 3 天的可用性
最近 72 小时过去 24 小时的可用性
当上游提供商发生错误时,如果您的请求筛选条件允许,我们可以通过路由到另一家健康的提供商来恢复服务。您可以通过以下接口以编程方式访问各提供商的正常运行时间数据: 端点 API。了解更多有关我们的 负载均衡和提供商路由选项。
基准测试
标准化评测得分,以及此模型针对共用基准测试提示词的实际输出。百分比越高越好,每项排名显示此模型在该来源所评测的模型中的位置。
| 来源 | 基准测试 | 得分 |
|---|---|---|
| Artificial Analysis | Mistral Large 4 (High) Intelligence Index | 26.1 |
| Artificial Analysis | Mistral Large 4 (High) Finance & Accounting Index | 39.1 |
| Artificial Analysis | Mistral Large 4 (High) Strategy & Ops Index | 43.3 |
| Artificial Analysis | Mistral Large 4 (High) Legal Index | 34.6 |
| Artificial Analysis | Mistral Large 4 (High) Healthcare & Medical Index | 36.7 |
| Artificial Analysis | Mistral Large 4 (High) Engineering Index | 33.4 |
| Artificial Analysis | Mistral Large 4 (High) Economics Index | 35.6 |
| Artificial Analysis | Mistral Large 4 (High) Cyber Index | 50.9 |
| Artificial Analysis | Mistral Large 4 (High) Openness Index | 38.9 |
| Artificial Analysis | Mistral Large 4 (High) AutomationBench-AA | 57.8% |
| Artificial Analysis | Mistral Large 4 (High) GDP.pdf | 17.6% |
| Artificial Analysis | Mistral Large 4 (High) CritPt | 10.0% |
| Artificial Analysis | Mistral Large 4 (High) HLE-Diamond | 17.3% |
| Artificial Analysis | Mistral Large 4 (High) Terminal-Bench 4.0 | 24.7% |
| Artificial Analysis | Mistral Large 4 (High) FrontierCode-AA 2.0 | 15.0% |
| Artificial Analysis | Mistral Large 4 (High) Terminal-Bench-Science | 4.3% |
| Artificial Analysis | Mistral Large 4 (High) CritPt v2 (Hard Subset) | 27.0% |
| Artificial Analysis | Mistral Large 4 (High) CritPt v2 | 50.3% |
| Artificial Analysis | Mistral Large 4 (High) AA-Omniscience Accuracy | 26.0% |
| Artificial Analysis | Mistral Large 4 (High) AA-Omniscience Non-Hallucination Rate | 71.8% |
| Artificial Analysis | Mistral Large 4 (High) CWE-Bench-AA | 58.3% |
| Artificial Analysis | Mistral Large 4 (High) DeepsecBench-AA | 15.0% |
| Artificial Analysis | Mistral Large 4 (High) CyberGym-E2E-AA | 79.4% |
| Artificial Analysis | Mistral Large 4 (High) MLCR-AA | 17.8% |
| Artificial Analysis | Mistral Large 4 (High) VIALS-AA | 20.5% |
| Artificial Analysis | Mistral Large 4 (Medium) Intelligence Index | 23.1 |
| Artificial Analysis | Mistral Large 4 (Medium) Finance & Accounting Index | 34.6 |
| Artificial Analysis | Mistral Large 4 (Medium) Strategy & Ops Index | 39.0 |
| Artificial Analysis | Mistral Large 4 (Medium) Legal Index | 31.3 |
| Artificial Analysis | Mistral Large 4 (Medium) Engineering Index | 30.9 |
| Artificial Analysis | Mistral Large 4 (Medium) Economics Index | 32.1 |
| Artificial Analysis | Mistral Large 4 (Medium) Cyber Index | 43.6 |
| Artificial Analysis | Mistral Large 4 (Medium) Openness Index | 38.9 |
| Artificial Analysis | Mistral Large 4 (Medium) AutomationBench-AA | 52.8% |
| Artificial Analysis | Mistral Large 4 (Medium) GDP.pdf | 16.8% |
| Artificial Analysis | Mistral Large 4 (Medium) CritPt | 11.1% |
| Artificial Analysis | Mistral Large 4 (Medium) HLE-Diamond | 13.9% |
| Artificial Analysis | Mistral Large 4 (Medium) Terminal-Bench 4.0 | 23.2% |
| Artificial Analysis | Mistral Large 4 (Medium) FrontierCode-AA 2.0 | 13.3% |
| Artificial Analysis | Mistral Large 4 (Medium) FrontierCode 1.1 | 37.8% |
| Artificial Analysis | Mistral Large 4 (Medium) Terminal-Bench-Science | 3.3% |
| Artificial Analysis | Mistral Large 4 (Medium) CritPt v2 (Hard Subset) | 22.5% |
| Artificial Analysis | Mistral Large 4 (Medium) CritPt v2 | 43.7% |
| Artificial Analysis | Mistral Large 4 (Medium) AA-Omniscience Accuracy | 22.5% |
| Artificial Analysis | Mistral Large 4 (Medium) AA-Omniscience Non-Hallucination Rate | 75.2% |
| Artificial Analysis | Mistral Large 4 (Medium) CWE-Bench-AA | 50.0% |
| Artificial Analysis | Mistral Large 4 (Medium) DeepsecBench-AA | 8.9% |
| Artificial Analysis | Mistral Large 4 (Medium) CyberGym-E2E-AA | 71.8% |
| Artificial Analysis | Mistral Large 4 Preview Intelligence Index | 25.3 |
| Artificial Analysis | Mistral Large 4 Preview Finance & Accounting Index | 38.1 |
| Artificial Analysis | Mistral Large 4 Preview Strategy & Ops Index | 43.4 |
| Artificial Analysis | Mistral Large 4 Preview Legal Index | 32.9 |
| Artificial Analysis | Mistral Large 4 Preview Healthcare & Medical Index | 33.8 |
| Artificial Analysis | Mistral Large 4 Preview Engineering Index | 30.9 |
| Artificial Analysis | Mistral Large 4 Preview Economics Index | 34.4 |
| Artificial Analysis | Mistral Large 4 Preview Cyber Index | 49.5 |
| Artificial Analysis | Mistral Large 4 Preview AutomationBench-AA | 59.9% |
| Artificial Analysis | Mistral Large 4 Preview GDP.pdf | 18.6% |
| Artificial Analysis | Mistral Large 4 Preview Harvey LAB-AA | 3.1% |
| Artificial Analysis | Mistral Large 4 Preview HLE | 35.0% |
| Artificial Analysis | Mistral Large 4 Preview AA-LCR | 81.3% |
| Artificial Analysis | Mistral Large 4 Preview GDPval-AA | 46.6% |
| Artificial Analysis | Mistral Large 4 Preview CritPt | 10.6% |
| Artificial Analysis | Mistral Large 4 Preview HLE-Diamond | 15.1% |
| Artificial Analysis | Mistral Large 4 Preview SciCode | 54.2% |
| Artificial Analysis | Mistral Large 4 Preview Terminal-Bench 4.0 | 26.8% |
| Artificial Analysis | Mistral Large 4 Preview FrontierCode-AA 2.0 | 13.6% |
| Artificial Analysis | Mistral Large 4 Preview FrontierCode 1.1 | 41.5% |
| Artificial Analysis | Mistral Large 4 Preview Terminal-Bench-Science | 4.8% |
| Artificial Analysis | Mistral Large 4 Preview CritPt v2 (Hard Subset) | 24.5% |
| Artificial Analysis | Mistral Large 4 Preview CritPt v2 | 43.7% |
| Artificial Analysis | Mistral Large 4 Preview AA-Omniscience Accuracy | 23.9% |
| Artificial Analysis | Mistral Large 4 Preview AA-Omniscience Non-Hallucination Rate | 58.7% |
| Artificial Analysis | Mistral Large 4 Preview CWE-Bench-AA | 50.8% |
| Artificial Analysis | Mistral Large 4 Preview DeepsecBench-AA | 16.0% |
| Artificial Analysis | Mistral Large 4 Preview CyberGym-E2E-AA | 81.7% |
| Artificial Analysis | Mistral Large 4 Preview MLCR-AA | 7.8% |
| Artificial Analysis | Mistral Large 4 Preview MMMU-Pro | 76.4% |
| Design Arena | Mistral Large 4 Preview Models Arena 3D Elo | 1231 |
| Design Arena | Mistral Large 4 Preview Models Arena Asciiart Elo | 1195 |
| Design Arena | Mistral Large 4 Preview Models Arena Code Categories Elo | 1247 |
| Design Arena | Mistral Large 4 Preview Models Arena Data Visualization Elo | 1272 |
| Design Arena | Mistral Large 4 Preview Models Arena Game Development Elo | 1198 |
| Design Arena | Mistral Large 4 Preview Models Arena SVG Elo | 1217 |
| Design Arena | Mistral Large 4 Preview Models Arena UI Component Elo | 1275 |
| Design Arena | Mistral Large 4 Preview Models Arena Website Elo | 1255 |
| Vals AI | Mistral Large 4 Preview BioMysteryBench | 67.0% |
| Vals AI | Mistral Large 4 Preview Code Migration | 30.6% |
| Vals AI | Mistral Large 4 Preview EMB v1.1 | 69.0% |
| Vals AI | Mistral Large 4 Preview Finance Agent v2.1 | 61.4% |
| Vals AI | Mistral Large 4 Preview Vals Finance Index | 56.9% |
| Vals AI | Mistral Large 4 Preview HLAB-Vals-1.1 | 17.5% |
| Vals AI | Mistral Large 4 Preview Immigration | 63.7% |
| Vals AI | Mistral Large 4 Preview IOI | 45.3% |
| Vals AI | Mistral Large 4 Preview Vals Legal Index | 65.6% |
| Vals AI | Mistral Large 4 Preview Legal Research Bench | 31.7% |
| Vals AI | Mistral Large 4 Preview MedCode | 40.7% |
| Vals AI | Mistral Large 4 Preview MedScribe | 80.4% |
| Vals AI | Mistral Large 4 Preview MysteryMechanism | 12.2% |
| Vals AI | Mistral Large 4 Preview ProofBench v1.1 | 10.0% |
| Vals AI | Mistral Large 4 Preview SWE-bench Multimodal | 33.1% |
| Vals AI | Mistral Large 4 Preview Tax Agent Bench | 30.4% |
| Vals AI | Mistral Large 4 Preview Terminal-Bench 4.0 | 27.8% |
| Vals AI | Mistral Large 4 Preview Vals Index | 48.7% |
| Vals AI | Mistral Large 4 Preview Vibe Code Bench 1-100 | 14.3% |
| Vals AI | Mistral Large 4 Preview Vibe Code Bench v1.1 | 78.4% |
应用
向此模型发送最多流量的公开应用。这能很好地反映实际生产工作负载的情况,也能提示此模型最适合哪些用例。
活动
此模型的 token 用量和请求流量随时间的变化。
API
可直接用于调用此模型的代码。OpenRouter 的 API 与 OpenAI 兼容——大多数 SDK 只需替换基础 URL 即可使用。切换模型时,唯一需要更改的是下方的模型标识符。
可通过欧盟区域内路由使用
Mistral: Mistral Large 4 可通过欧盟区域内路由使用。发送至 eu.orr.api.report.finance 的请求会在欧盟境内解密,并且仅路由至欧洲(欧盟/欧洲经济区、英国和瑞士)的共享提供商端点。
- 欧洲的提供商
- Mistral
- 欧洲价格
- Mistral:输入 US$0.77/M tokens,输出 US$2.31/M tokens
- 基础 URL
https://eu.orr.api.report.finance/api/v1- 套餐
- 需要 Business 或 Enterprise 套餐。Business 支持自助使用:无需合同,无每月最低消费。 用于欧盟区域内路由的 OpenRouter Business
故障时关闭:如果欧洲没有可用的共享提供商端点,请求将报错,而不会离开该区域。使用您自己的 BYOK 密钥处理的请求会发送至您部署所在的区域,该区域由您控制。 区域内路由文档
文档
在 OpenRouter 上使用 Mistral: Mistral Large 4 构建应用的指南、实用教程、博客文章和工具。
常见问题
Mistral Large 4 is a frontier multimodal (text and image input) model from Mistral AI built for reasoning, coding, and agentic workloads. It offers a 1M-token context window with up to 256K output tokens, and supports tool calling and structured outputs.
OpenRouter 上的 Mistral Large 4 文档包括 How to Benchmark LLMs 博客文章和How to Stop an AI Agent Stuck in an Infinite Loop 博客文章。快速入门介绍了 API 密钥和您的首次请求,API 参考列出了所有参数。
Mistral Large 4 的费用为 US$0.70/百万输入 token 和 US$2.10/百万输出 token,Cache Read,价格为 US$0.07/M tokens 单独计费。
Mistral Large 4 的上下文窗口为 1,048,576 个 token。它最多支持 262,144 个补全 token。
是的。Mistral Large 4 接受 tools 和 tool_choice,以进行函数调用。 它还支持通过 response_format 中的 JSON schema 实现结构化输出。
Mistral Large 4 接受 文本和图像 作为输入,并返回 文本。
在 OpenRouter 上,Mistral Large 4 由 1 家提供商提供服务:Mistral。
Mistral Medium 3.5, Mistral Small 4, Devstral 2 2512 及另外 16 个 是 Mistral 的其他 文本 模型。
Mistral Large 4 于 2026年10月6日 发布。