1. 图表数据
  2. arena
  3. Claude 3.7 Sonnet Thinking shows strength across a variety o
arena表格图 · 150 个数据点 · Claude 3

Claude 3.7 Sonnet Thinking shows strength across a variety of categories

arena 发布的一张图背后的数字,用原生图表重画。对照原图,研究数据,再做你自己的。

图表来自 arena. 来源: x.com

简化重绘

由我们用下面的数字重新绘制。悬停、点按或用方向键查看。

这是简化重绘:它用我们自己的样式展示数字,不是原图的样子。强调色标出这张图的主角。表格里带 ~ 的是估算值。

数字

150 个点中前 10 个免费
分组系列数值来源置信
gpt-4.5-preview-2025-02-27Overall~1估算0.50
chatgpt-4o-latest-20250129Overall~3估算0.50
grok-3-preview-02-24Overall~1估算0.50
gemini-2.0-pro-exp-02-05Overall~3估算0.50
o1-2024-12-17Overall~6估算0.50
claude-3-7-sonnet-20250219-thinking-32kOverall~14估算0.50
deepseek-r1Overall~6估算0.50
gemini-2.0-flash-thinking-exp-01-21Overall~3估算0.50
o1-previewOverall~9估算0.50
claude-3-7-sonnet-20250219Overall~16估算0.50

Max 数据,目前留邮箱免费看

查看全部 150 个数据点并下载文件

这张图的完整表格、JSON、CSV 和 agent skill 文件。它们属于 Max;原型阶段,留邮箱即可解锁。

解锁全部 35 家巨头。每周一封,随时退订。 看看能得到什么 · 登录

gemini-2.0-flash-001Overall~6估算0.50
gemma-3-27b-itOverall~8估算0.50
qwen2.5-maxOverall~9估算0.50
claude-3-5-sonnet-20241022Overall~27估算0.50
deepseek-v3Overall~12估算0.50
gpt-4.5-preview-2025-02-27Overall w/ Style Control~1估算0.50
chatgpt-4o-latest-20250129Overall w/ Style Control~2估算0.50
grok-3-preview-02-24Overall w/ Style Control~3估算0.50
gemini-2.0-pro-exp-02-05Overall w/ Style Control~3估算0.50
o1-2024-12-17Overall w/ Style Control~3估算0.50
claude-3-7-sonnet-20250219-thinking-32kOverall w/ Style Control~4估算0.50
deepseek-r1Overall w/ Style Control~4估算0.50
gemini-2.0-flash-thinking-exp-01-21Overall w/ Style Control~6估算0.50
o1-previewOverall w/ Style Control~7估算0.50
claude-3-7-sonnet-20250219Overall w/ Style Control~7估算0.50
gemini-2.0-flash-001Overall w/ Style Control~11估算0.50
gemma-3-27b-itOverall w/ Style Control~11估算0.50
qwen2.5-maxOverall w/ Style Control~11估算0.50
claude-3-5-sonnet-20241022Overall w/ Style Control~11估算0.50
deepseek-v3Overall w/ Style Control~12估算0.50
gpt-4.5-preview-2025-02-27Hard Prompts~1估算0.50
chatgpt-4o-latest-20250129Hard Prompts~5估算0.50
grok-3-preview-02-24Hard Prompts~5估算0.50
gemini-2.0-pro-exp-02-05Hard Prompts~3估算0.50
o1-2024-12-17Hard Prompts~3估算0.50
claude-3-7-sonnet-20250219-thinking-32kHard Prompts~5估算0.50
deepseek-r1Hard Prompts~4估算0.50
gemini-2.0-flash-thinking-exp-01-21Hard Prompts~3估算0.50
o1-previewHard Prompts~6估算0.50
claude-3-7-sonnet-20250219Hard Prompts~15估算0.50
gemini-2.0-flash-001Hard Prompts~5估算0.50
gemma-3-27b-itHard Prompts~13估算0.50
qwen2.5-maxHard Prompts~5估算0.50
claude-3-5-sonnet-20241022Hard Prompts~15估算0.50
deepseek-v3Hard Prompts~15估算0.50
gpt-4.5-preview-2025-02-27Hard Prompts w/ Style Control~1估算0.50
chatgpt-4o-latest-20250129Hard Prompts w/ Style Control~3估算0.50
grok-3-preview-02-24Hard Prompts w/ Style Control~3估算0.50
gemini-2.0-pro-exp-02-05Hard Prompts w/ Style Control~2估算0.50
o1-2024-12-17Hard Prompts w/ Style Control~2估算0.50
claude-3-7-sonnet-20250219-thinking-32kHard Prompts w/ Style Control~1估算0.50
deepseek-r1Hard Prompts w/ Style Control~3估算0.50
gemini-2.0-flash-thinking-exp-01-21Hard Prompts w/ Style Control~5估算0.50
o1-previewHard Prompts w/ Style Control~5估算0.50
claude-3-7-sonnet-20250219Hard Prompts w/ Style Control~11估算0.50
gemini-2.0-flash-001Hard Prompts w/ Style Control~2估算0.50
gemma-3-27b-itHard Prompts w/ Style Control~15估算0.50
qwen2.5-maxHard Prompts w/ Style Control~7估算0.50
claude-3-5-sonnet-20241022Hard Prompts w/ Style Control~15估算0.50
deepseek-v3Hard Prompts w/ Style Control~16估算0.50
gpt-4.5-preview-2025-02-27Coding~1估算0.50
chatgpt-4o-latest-20250129Coding~1估算0.50
grok-3-preview-02-24Coding~1估算0.50
gemini-2.0-pro-exp-02-05Coding~3估算0.50
o1-2024-12-17Coding~3估算0.50
claude-3-7-sonnet-20250219-thinking-32kCoding~7估算0.50
deepseek-r1Coding~1估算0.50
gemini-2.0-flash-thinking-exp-01-21Coding~3估算0.50
o1-previewCoding~4估算0.50
claude-3-7-sonnet-20250219Coding~8估算0.50
gemini-2.0-flash-001Coding~4估算0.50
gemma-3-27b-itCoding~15估算0.50
qwen2.5-maxCoding~6估算0.50
claude-3-5-sonnet-20241022Coding~7估算0.50
deepseek-v3Coding~14估算0.50
gpt-4.5-preview-2025-02-27Math~1估算0.50
chatgpt-4o-latest-20250129Math~14估算0.50
grok-3-preview-02-24Math~1估算0.50
gemini-2.0-pro-exp-02-05Math~2估算0.50
o1-2024-12-17Math~1估算0.50
claude-3-7-sonnet-20250219-thinking-32kMath~1估算0.50
deepseek-r1Math~3估算0.50
gemini-2.0-flash-thinking-exp-01-21Math~2估算0.50
o1-previewMath~2估算0.50
claude-3-7-sonnet-20250219Math~12估算0.50
gemini-2.0-flash-001Math~2估算0.50
gemma-3-27b-itMath~6估算0.50
qwen2.5-maxMath~3估算0.50
claude-3-5-sonnet-20241022Math~14估算0.50
deepseek-v3Math~14估算0.50
gpt-4.5-preview-2025-02-27Creative Writing~1估算0.50
chatgpt-4o-latest-20250129Creative Writing~1估算0.50
grok-3-preview-02-24Creative Writing~1估算0.50
gemini-2.0-pro-exp-02-05Creative Writing~1估算0.50
o1-2024-12-17Creative Writing~6估算0.50
claude-3-7-sonnet-20250219-thinking-32kCreative Writing~6估算0.50
deepseek-r1Creative Writing~2估算0.50
gemini-2.0-flash-thinking-exp-01-21Creative Writing~1估算0.50
o1-previewCreative Writing~9估算0.50
claude-3-7-sonnet-20250219Creative Writing~6估算0.50
gemini-2.0-flash-001Creative Writing~2估算0.50
gemma-3-27b-itCreative Writing~7估算0.50
qwen2.5-maxCreative Writing~4估算0.50
claude-3-5-sonnet-20241022Creative Writing~18估算0.50
deepseek-v3Creative Writing~17估算0.50
gpt-4.5-preview-2025-02-27Instruction Following~1估算0.50
chatgpt-4o-latest-20250129Instruction Following~3估算0.50
grok-3-preview-02-24Instruction Following~1估算0.50
gemini-2.0-pro-exp-02-05Instruction Following~2估算0.50
o1-2024-12-17Instruction Following~3估算0.50
claude-3-7-sonnet-20250219-thinking-32kInstruction Following~3估算0.50
deepseek-r1Instruction Following~6估算0.50
gemini-2.0-flash-thinking-exp-01-21Instruction Following~2估算0.50
o1-previewInstruction Following~8估算0.50
claude-3-7-sonnet-20250219Instruction Following~11估算0.50
gemini-2.0-flash-001Instruction Following~6估算0.50
gemma-3-27b-itInstruction Following~8估算0.50
qwen2.5-maxInstruction Following~8估算0.50
claude-3-5-sonnet-20241022Instruction Following~7估算0.50
deepseek-v3Instruction Following~13估算0.50
gpt-4.5-preview-2025-02-27Longer Query~1估算0.50
chatgpt-4o-latest-20250129Longer Query~1估算0.50
grok-3-preview-02-24Longer Query~1估算0.50
gemini-2.0-pro-exp-02-05Longer Query~1估算0.50
o1-2024-12-17Longer Query~3估算0.50
claude-3-7-sonnet-20250219-thinking-32kLonger Query~3估算0.50
deepseek-r1Longer Query~4估算0.50
gemini-2.0-flash-thinking-exp-01-21Longer Query~2估算0.50
o1-previewLonger Query~7估算0.50
claude-3-7-sonnet-20250219Longer Query~6估算0.50
gemini-2.0-flash-001Longer Query~7估算0.50
gemma-3-27b-itLonger Query~7估算0.50
qwen2.5-maxLonger Query~6估算0.50
claude-3-5-sonnet-20241022Longer Query~6估算0.50
deepseek-v3Longer Query~7估算0.50
gpt-4.5-preview-2025-02-27Multi-Turn~1估算0.50
chatgpt-4o-latest-20250129Multi-Turn~2估算0.50
grok-3-preview-02-24Multi-Turn~1估算0.50
gemini-2.0-pro-exp-02-05Multi-Turn~2估算0.50
o1-2024-12-17Multi-Turn~7估算0.50
claude-3-7-sonnet-20250219-thinking-32kMulti-Turn~9估算0.50
deepseek-r1Multi-Turn~2估算0.50
gemini-2.0-flash-thinking-exp-01-21Multi-Turn~4估算0.50
o1-previewMulti-Turn~7估算0.50
claude-3-7-sonnet-20250219Multi-Turn~7估算0.50
gemini-2.0-flash-001Multi-Turn~5估算0.50
gemma-3-27b-itMulti-Turn~7估算0.50
qwen2.5-maxMulti-Turn~8估算0.50
claude-3-5-sonnet-20241022Multi-Turn~11估算0.50
deepseek-v3Multi-Turn~8估算0.50
下载这张图的数据

Max 订阅者可以通过 API 取同样的数据。 API 文档

API · Max

用 API 调用这张图

Max 套餐可以取这张图的数字(每个数字都带来源),格式为 JSON 或 CSV,附重画规格;也可以让智能体通过 MCP 调用。图表 id:ad22ff385c58。

路径返回套餐
GET/api/v1/charts/ad22ff385c58标题、发布者、日期、图型、数据点数;用 Max 密钥时附数据链接开放
GET/api/v1/charts/ad22ff385c58/data每个数据点及其来源和置信度,加上重画规格。加 ?format=csv 得到 CSVMax
GET/api/v1/charts/ad22ff385c58/render?view=1本页使用的只读绘图视图开放
POST/mcp工具 get_chart_data 加这个 id,供 Claude、Cursor 等智能体使用Max
curlMax
curl -s https://viral-chart.com/api/v1/charts/ad22ff385c58/data \
  -H "Authorization: Bearer $VIRALCHART_KEY"
Claude CodeMax
claude mcp add --transport http viralchart https://viral-chart.com/mcp \
  --header "Authorization: Bearer $VIRALCHART_KEY"

每个密钥每分钟 60 次,超出返回 429 和 Retry-After。每个账号最多 5 个密钥,在账户页创建,只显示一次。没有 Max 时数据接口返回 402,并附套餐链接。这些数字描述的是发布者拥有的图,每个文件都带授权说明。

做一张这样的图

这张图的形式,你的数字。

实时预览。所有内容都留在你的设备上。

选择形式

简化形式:用我们自己的样式,不是原图的样子。你的数字从这张图的数据开始。

Anthropic 的更多数据

全部图表数据

常见问题

这些数字有多准?

它们由两遍独立读图估算得出,置信度 0.50(0 到 1)。图上印着的数字读得准,从坐标轴读出的值可能有几个百分点的偏差。

这些数据可以直接用吗?

这些数字描述的是 arena 发布的图表。图片、版式和样式归他们所有。使用数字时请注明发布者,图片请用你自己的数据来做。

怎么拿到文件?

JSON、CSV 和 agent skill 文件在本页的邮箱解锁区。原型里它们是静态文件:/api/charts/ad22ff385c58.json 和 /api/charts/ad22ff385c58.skill.md。