Claude Sonnet 5

Anthropic·5 个数据来源·更新于 2026-11-10

上下文窗口1M厂商官网
最大输出128K厂商官网
输入 / 百万 Token$2厂商官网
输出 / 百万 Token$10厂商官网

评测表现

只在同一来源、同一口径下比较。切换来源可查看不同测试结果。

4 个评测来源

当前配置:各基准设置见下表

采集于 2026-11-10查看原始页面
LLM Stats · 公开基准成绩
暂无综合分数当前所选配置的来源分数

不同来源的分数不能直接相加或横向比较;评测任务成本不是 API Token 单价。

逐项基准成绩

24 项已收录成绩 · 保留原始单位和测试条件

ArXivMath厂商自报 · 越高越好
72.2%

Research-level final-answer math (MathArena, April + May 2026 releases, 81 problems). With tools: 72.2%. Without tools: 65.7%. Extended thinking, averaged over 4 attempts per problem.

成绩出处
AutomationBench厂商自报 · 越高越好
13.5%

Zapier end-to-end business-workflow benchmark, private held-out leaderboard set. Max effort.

成绩出处
BenchCAD厂商自报 · 越高越好
37.3%

BenchCAD Vision2Code voxel IoU. With Python tools: 0.373. Without tools: 0.266. Random 1,000-file subset averaged over 5 runs.

成绩出处
BrowseComp厂商自报 · 越高越好
84.7%

Agentic search. Single-agent with web search, web fetch, code execution, adaptive thinking at max effort, 10M-token limit, context compaction at 200k tokens. Multi-agent reaches 86.6%.

成绩出处
ChartMuseum厂商自报 · 越高越好
86.7%

Chart QA requiring visual reasoning (1,162 questions). With Python tools: 86.7%. Without tools: 70.1%. Test split averaged over 5 runs, Claude Sonnet 4.6 grader.

成绩出处
CharXiv-R厂商自报 · 越高越好
88.3%

CharXiv Reasoning (1,000 validation questions). With Python tools: 88.3%. Without tools: 77.0%. Averaged over 5 runs, Claude Sonnet 4.6 grader.

成绩出处
DeepSWE 1.1来源汇总 · 越高越好
54%

DeepSWE v1.1 leaderboard, mini-swe-agent harness, max effort; Pass@1 54% ± 4%.

成绩出处
FrontierCode厂商自报 · 越高越好
38.8%

FrontierCode v1, Cognition's production-quality coding evaluation, at max effort.

成绩出处
FrontierCode 1.1来源汇总 · 越高越好
42.7%

FrontierCode 1.1 current leaderboard; mergeability score at xhigh effort: 42.7%.

成绩出处
GDP.pdf厂商自报 · 越高越好
81.6%

Expert multimodal document QA (Surge AI, 100 prompts/PDFs). Mean criteria pass rate with Python tools: 81.6%. Without tools: 67.5%. Internal harness, averaged over 5 runs.

成绩出处
HealthBench Professional厂商自报 · 越高越好
57.8%

Length-adjusted score with adaptive thinking at max effort. Claude Opus 4.8 grader, averaged over 5 trials.

成绩出处
Humanity's Last Exam厂商自报 · 越高越好
57.4%

Multidisciplinary reasoning (2,500 questions). With tools (web search, web fetch, code execution): 57.4%. Without tools: 43.2%. Auto thinking, 1M total-token cap, Claude Opus 4.6 grader.

成绩出处
Legal Agent Benchmark厂商自报 · 越高越好
5.8%

All-pass rate on Harvey's held-out set (mean criterion-pass rate 91.2%). Full public set all-pass rate: 8.9% (88.26% mean criterion-pass). Adaptive thinking at max effort.

成绩出处
LifeSci Diamond (2026-11)来源汇总 · 越高越好
1.8%

Strict pass rate on 55 regulatory-affairs tasks. Plain API via Vercel AI Gateway with a web_search tool (Exa), provider default settings, one answer per task, graded by GPT-6.1 Sol at medium reasoning effort (Arca, 4 Nov 2026 edition).

成绩出处
OfficeQA Pro厂商自报 · 越高越好
59.4%

Grounded reasoning over U.S. Treasury Bulletin documents (133-question subset). Internal agentic harness with extracted-text documents and code-execution tools. Exact-match, mean of 5 trials. OfficeQA full set: 73.3%.

成绩出处
OSWorld-Verified厂商自报 · 越高越好
81.2%

Agentic computer use. 361 tasks at 100 steps, 1080p resolution, adaptive thinking at max effort. First-attempt success rate averaged over 5 runs. Updated harness (zoom-tool bug fix, 128K max tokens per turn).

成绩出处
SWE-bench Multilingual厂商自报 · 越高越好
78.3%

300 problems across 9 programming languages. Averaged over 5 trials at max effort.

成绩出处
SWE-Bench Multimodal厂商自报 · 越高越好
28.1%

Visual context (screenshots, design mockups) added to issue descriptions. Internal harness, averaged over 5 trials.

成绩出处
SWE-Bench Pro厂商自报 · 越高越好
63.2%

Harder SWE-bench variant: actively-maintained repos with larger multi-file diffs and reduced public ground-truth leakage. Averaged over 5 trials at max effort.

成绩出处
SWE-Bench Verified厂商自报 · 越高越好
85.2%

Agentic coding (SWE-bench Verified, 500-problem subset). Standard configuration, adaptive thinking at max effort, thinking blocks included in sampling, averaged over 5 trials.

成绩出处
Terminal-Bench 2.0厂商自报 · 越高越好
80.4%

Agentic terminal coding (Terminal-Bench 2.1) with the mini-SWE-agent harness at xhigh effort. Mean reward over 5 attempts across 89 unique tasks (445 trials).

成绩出处
Terminal-Bench 4.0来源汇总 · 越高越好
12.4%

Terminal-Bench 4.0 public leaderboard; resolution rate 12.4% ± 3.1%; Claude Code agent; max effort.

成绩出处
Toolathlon厂商自报 · 越高越好
54.3%

Pass@1 averaged over 3 trials across all 108 tasks. Internal harness with adaptive thinking at max effort. Pass@3: 63.0%, Pass^3: 40.7%.

成绩出处
USAMO 2026厂商自报 · 越高越好
33.39分

Proof-based math (USAMO 2026, held March 21-22 2026). MathArena grading methodology (proofs rewritten by a neutral model and judged by a 3-model panel; minimum score taken). High effort, 300k token limit, averaged over 10 attempts. 79.5%.

成绩出处

技术规格

只展示有可靠来源的资料;优先采用厂商官网与官方模型仓库。

发布日期
2026-06-30厂商官网
知识截止
2026-01厂商官网
上下文窗口
1M Token厂商官网
最大输入
1M TokenLLM Stats
最大输出
128K Token厂商官网
输入模态
文本、图像厂商官网
输出模态
文本厂商官网
推理模式
支持厂商官网
许可证
ProprietaryLLM Stats

价格与用量

美元 / 百万 Token。不同服务商或测试配置的报价分别列出。

模型 API 价格,单位为美元每百万 Token
来源 / 服务商输入输出缓存读取
厂商官网 · Claude Sonnet 5$2$10$0.2
Anthropic$2$10$0.2
DeepInfra$3$15待补充

估算一次用量

按表格第一条报价计算,普通文本、无缓存。

预计费用$3
不含工具、图片、推理额外计费或阶梯价格;以服务商实际账单为准。

每个数字,都有出处

同名模型通过开发机构和明确版本关联。预览版、日期版本和不同规模模型分别建档。

厂商官网

厂商官方文档 · 基础规格与标准 API 报价

1 条记录最近采集 2026-11-10

LLM Stats

公开模型页与公开榜单;自报成绩单独标识

1 条记录最近采集 2026-11-10

BenchLM

公开评测数据;保留来源原始口径

1 条记录最近采集 2026-11-10

Arena

Arena 官方公开数据集 · CC BY 4.0

5 条记录最近采集 2026-11-10

OpenRouter

公开模型目录与 API 报价

1 条记录最近采集 2026-11-10
查看全部配置与采集记录
来源 / 配置采集时间原始数据
LLM StatsClaude Sonnet 5SHA-256 38986ec9b954a818…2026-11-10T13:53:09Z查看来源
BenchLMClaude Sonnet 52026-11-10查看来源
ArenaClaude Sonnet 5 (High)2026-11-10T14:07:52Z查看来源
ArenaClaude Sonnet 5 (High)2026-11-10T14:07:52Z查看来源
ArenaClaude Sonnet 5 (High)2026-11-10T14:07:52Z查看来源
ArenaClaude Sonnet 5 (High)2026-11-10T14:07:52Z查看来源
ArenaClaude Sonnet 5 (High)2026-11-10T14:07:52Z查看来源
OpenRouterClaude Sonnet 5SHA-256 63cbc45064d84060…2026-11-10T13:51:49Z查看来源
厂商官网Claude Sonnet 5SHA-256 188e8eb2ba29b33c…2026-11-10T14:06:47Z查看来源
继续浏览模型排行榜