DataLearner 标志

Claude Opus 4.8vsGemini 3.1 Pro Preview

在 13 个共同 benchmark 中，Claude Opus 4.8 整体领先：Claude Opus 4.8 领先 8 项，Gemini 3.1 Pro Preview 领先 5 项，持平 0 项，平均分差 +12.04。

Anthropic
Claude Opus 4.8

Anthropic · 2026-05-28 · 推理大模型

Google Deep Mind
Gemini 3.1 Pro Preview

Google Deep Mind · 2026-02-20 · 多模态大模型

Claude Opus 4.88 项(62%)(38%)5 项Gemini 3.1 Pro Preview

评测分数

按能力类目分组，每组内按分差大小排列；共 13 项。

Coding and Software Engineer

Claude Opus 4.8 领先 4/5
评测项Claude Opus 4.8Gemini 3.1 Pro Preview分差
Text Arena (Coding)1,5459 / 35Normal (No Tools)1,46123 / 35Normal (No Tools)+83.56
DeepSWE5913 / 27Deep Thinking (With Tools)1227 / 27Thinking High (With Tools)+47
SWE-Bench Pro - Public69.204 / 57Extended (with tools)54.2034 / 57Thinking High (With Tools)+15
SWE-bench Verified88.605 / 114Extended (with tools)80.6011 / 114Thinking High (With Tools)+8
WeirdML v270.4518 / 52Normal (With Tools)72.1017 / 52Normal (With Tools)-1.65

AI Agent - Tool Usage

Claude Opus 4.8 领先 3/3
评测项Claude Opus 4.8Gemini 3.1 Pro Preview分差
OSWorld-Verified83.404 / 26Extended (with tools)76.2012 / 26Thinking (With Tools)+7.20
Terminal-Bench 2.178.9020 / 44Thinking High (With Tools)73.8028 / 44Thinking High (With Tools)+5.10
MCP-Atlas82.206 / 38Deep Thinking (With Tools)78.2012 / 38Thinking High (With Tools)+4

General Knowledge

胶着 2/2
评测项Claude Opus 4.8Gemini 3.1 Pro Preview分差
HLE57.908 / 181Extended (with tools)51.4024 / 181Thinking High (With Tools)+6.50
LiveBench78.794 / 115Deep Thinking (No Tools)79.933 / 115Thinking High (No Tools)-1.14

AI Agent - Information Search

Gemini 3.1 Pro Preview 领先 1/1
评测项Claude Opus 4.8Gemini 3.1 Pro Preview分差
BrowseComp84.309 / 54Thinking High (With Tools + Internet)85.905 / 54Thinking High (With Tools + Internet)-1.60

常识推理

Gemini 3.1 Pro Preview 领先 1/1
评测项Claude Opus 4.8Gemini 3.1 Pro Preview分差
SimpleBench64.8010 / 67Normal (No Tools)79.602 / 67Normal (No Tools)-14.80

科学与综合推理

Gemini 3.1 Pro Preview 领先 1/1
评测项Claude Opus 4.8Gemini 3.1 Pro Preview分差
GPQA Diamond93.6011 / 226Thinking High (No Tools)94.304 / 226Thinking High (No Tools)-0.70

规格对比

字段Claude Opus 4.8Gemini 3.1 Pro Preview
发布机构AnthropicGoogle Deep Mind
发布时间2026-05-282026-02-20
模型类型推理大模型多模态大模型
架构稠密模型稠密模型
参数规模暂无数据暂无数据
上下文长度1M1M
最大输出125K64K

API 调用价格

价格优先使用 DataLearner 配置的 API 记录；缺失项不做推测。

价格项Claude Opus 4.8Gemini 3.1 Pro Preview
文本输入$5 / 1M tokens$2 / 1M tokens
文本输出$25 / 1M tokens$12 / 1M tokens
缓存读取$0.5 / 1M tokens暂无公开价格
缓存写入$6.25 / 1M tokens暂无公开价格

小结

  • Claude Opus 4.8在以下类目领先:Coding and Software Engineer (4/5)、AI Agent - Tool Usage (3/3)
  • Gemini 3.1 Pro Preview在以下类目领先:AI Agent - Information Search (1/1)、常识推理 (1/1)、科学与综合推理 (1/1)
  • 胶着类目:General Knowledge

13 个共同 benchmark 上，Claude Opus 4.8 平均高出 12.04 分。

单项差距最大的 benchmark：Text Arena (Coding) — Claude Opus 4.8 1,545，Gemini 3.1 Pro Preview 1,461（分差 +83.56）。

本页正文由结构化模型、价格与 benchmark 数据生成，不使用实时 LLM 撰写。

Claude Opus 4.8 详情Gemini 3.1 Pro Preview 详情·在工具里自定义对比