开始使用
API 接入
Claude API 模型说明
发布页给出的模型 ID 为 claude-opus-4-8；以当前平台权限与价格为准。
Merge Gateway API文档
目录部署标识：anthropic/claude-opus-4-8。具体区域、权限与价格以服务商为准。
Tempr Gateway API文档
目录部署标识：anthropic/claude-opus-4-8。具体区域、权限与价格以服务商为准。
NEAR AI Cloud API文档
目录部署标识：anthropic/claude-opus-4-8。具体区域、权限与价格以服务商为准。
Impossibl API文档
目录部署标识：anthropic/claude-opus-4-8。具体区域、权限与价格以服务商为准。
Eden AI API文档
目录部署标识：anthropic/claude-opus-4-8。具体区域、权限与价格以服务商为准。
LLM Gateway API文档
目录部署标识：anthropic/claude-opus-4-8。具体区域、权限与价格以服务商为准。
CrossModel API文档
目录部署标识：anthropic/claude-opus-4-8。具体区域、权限与价格以服务商为准。
Anthropic API文档
目录部署标识：claude-opus-4-8。具体区域、权限与价格以服务商为准。
公开成绩与计算依据
综合参考分根据已公开的评测资料整理。查看发布日期依据
综合参考分 18.7 · 14项基准
推理求解 10.5
FrontierMath · 56.1 %
Claude Opus 4.8 · Published page openai56; row FrontierMath Tier 4 (v2); preserve page footnotes and task-specific settings · 2026-07-09公开
GPQA Diamond · 92%
openai.com · 官方发布比较 · Diamond
Claude Opus 4.8 · Official published comparison; GPQA Diamond; exact column Claude Opus 4.8. Footnotes and all competing runs retained in source. · 2026-07-09公开
Terminal-Bench-Science · 4.29%
Vals AI · 独立评测 · 0.1.0
Claude Opus 4.8 / Mini-SWE-agent · Current displayed evaluator row. Mini-SWE-agent. Scores are not pooled across harnesses; source methodology and repeat counts retained.
工具与任务执行 41.3
APEX-Agents · 39.4
huggingface.co · 官方发布比较 · current published task suite
Claude Opus 4.8 (max) · Official published comparison; APEX-Agents; exact column Claude Opus 4.8 (max). Footnotes and all competing runs retained in source.
Terminal-Bench · 23.23%
Vals AI · 独立评测 · 4.0
Claude Opus 4.8 / Mini-SWE-agent · Current displayed evaluator row. Mini-SWE-agent. Scores are not pooled across harnesses; source methodology and repeat counts retained.
AutomationBench · 15.5 %
openai.com · 官方发布比较 · current published task suite
Claude Opus 4.8 · Published page openai56; row AutomationBench; preserve page footnotes and task-specific settings · 2026-07-09公开
AA-Briefcase · 0.38686868686868686 比例
Artificial Analysis · 独立评测 · 1.1
Claude Opus 4.8 (Max) · aa-briefcase-rubric-reviewed-method-20261029
MCP Atlas · 82.2 %
Scale AI · 独立评测 · 1000 tasks
claude-opus-4-8 (max) · scale-mcp-reviewed-method-20261029
编程 0.0
DeepSWE · 59.0 %
openai.com · 官方发布比较 · 1.1
Claude Opus 4.8 · Published page openai56; row DeepSWE v1.1; preserve page footnotes and task-specific settings · 2026-07-09公开
语言与信息处理 26.0
IFBench · 62.2
huggingface.co · 官方发布比较 · current published task suite
Opus 4.8 · Published page qwen38-weights; row IFBench; preserve page footnotes and task-specific settings
GDP.pdf · 22.5 %
openai.com · 官方发布比较 · current published task suite
Claude Opus 4.8 · Published page openai56; row gdp.pdf; preserve page footnotes and task-specific settings · 2026-07-09公开
SimpleQA Verified · 0.53 比例
Epoch AI · 独立评测 · run 1.2.0
claude-opus-4-8_max · epoch-simpleqa-reviewed-method-20261029-task-version-1.2.0 · 2026-08-27公开
多模态理解 5.6
Humanity Sixth Sense · 23.5 %
Scale AI · 独立评测 · current published task suite
Claude-Opus-4.8 (max) · scale-hss-reviewed-method-20261029
MMMU-Pro · 78.9 / 82.7
huggingface.co · 官方发布比较 · Standard 10
Claude Opus 4.8 (max) · Published page kimi-k3; row MMMU-Pro; preserve page footnotes and task-specific settings
资料来源
- 版本日期来源 · 版本资料 · 2026-10-29查阅
- openai.com · official_release_comparison · 2026-10-29查阅
- Vals AI · independent_evaluation · 2026-10-29查阅
- huggingface.co · official_release_comparison · 2026-10-29查阅
- Vals AI · independent_evaluation · 2026-10-29查阅
- Artificial Analysis · independent_evaluation · 2026-10-29查阅
- Scale AI · independent_evaluation · 2026-10-29查阅
- huggingface.co · official_release_comparison · 2026-10-29查阅
- Epoch AI · independent_evaluation · 2026-10-29查阅
- Scale AI · independent_evaluation · 2026-10-29查阅