See key specs and per-benchmark scores for each model/mode. Scroll horizontally for all columns. 当前对比 2 个模型的评测数据与核心参数。
Opus 4.5
Anthropic
Claude Opus 4.6
Anthropic
Each axis is a category average, normalized to a 100-point radar.
Relative edge: 编程与软件工程 +5.5 / Relative gap: 数学推理 -18.1
Relative edge: 数学推理 +18.1 / Relative gap: 编程与软件工程 -5.5
Method: for each model and benchmark, the chart first averages all scores in the current mode scope instead of taking the best score, then averages those benchmark scores within each category. Only benchmarks with at least two selected models scored are included; missing values are not counted as zero.
Best overall
Claude Opus 4.6 · 74.06
Best single
Claude Opus 4.6 · τ²-Bench - Telecom 99.25
Modality coverage
Opus 4.5 · 2 modalities
Head to head
17
Benchmarks
3
Wins
14
Losses
-8.99
Average diff
Compare benchmark results across thinking modes and tool usage.
Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology
Complete scores for each model/mode across selected benchmarks.
17 benchmarks with comparable scores. Each model shows its best score; mode label is displayed below.
| Benchmark | Opus 4.5 | Claude Opus 4.6 |
|---|---|---|
ARC-AGI 综合评估 | 80.00Extended Thinking | 92.00Extended Thinking |
ARC-AGI-2 综合评估 | 37.60Extended Thinking | 66.30Extended Thinking |
GPQA Diamond 综合评估 | 87.00Extended Thinking | 91.31Extended Thinking |
HLE 综合评估 | 43.20Extended Thinking ｜ Tools | 53.00Extended Thinking ｜ Tools |
LiveBench 综合评估 | 75.9664K | 76.33Thinking Level · High |
LiveCodeBench 编程与软件工程 | 87.00Extended Thinking ｜ Tools | 76.00Extended Thinking |
SWE-bench Verified 编程与软件工程 | 80.90Extended Thinking ｜ Tools | 80.84Extended Thinking ｜ Tools |
MMMU 多模态理解 | 80.70Extended Thinking | 77.30Extended Thinking ｜ Tools |
Simple Bench 常识推理 | 62.00Extended Thinking | 67.60Standard Mode |
τ²-Bench Agent能力评测 | 81.99Extended Thinking ｜ Tools | 91.89Extended Thinking ｜ Tools |
τ²-Bench - Telecom Agent能力评测 | 90.70Extended Thinking ｜ Tools | 99.25Extended Thinking ｜ Tools |
FrontierMath 数学推理 | 20.70Extended Thinking | 40.70Thinking Level · High |
4.20Standard Mode | 22.90Thinking Level · High | |
IF Bench 指令跟随 | 58.00Extended Thinking ｜ Tools | 94.00Extended Thinking |
MCP-Atlas AI Agent - 工具使用 | 69.80Thinking Level · High ｜ Tools | 76.80Deep Thinking Mode ｜ Tools |
Terminal Bench 2.0 AI Agent - 工具使用 | 59.30Extended Thinking ｜ Tools | 65.40Extended Thinking ｜ Tools |
Pinch Bench OpenClaw智能体能力综合测评 | 87.20Extended Thinking ｜ Tools | 87.40Thinking Enabled ｜ Tools |
Side-by-side input/output token pricing
Licensing, MoE architecture, and multi-modality support.
| Features & specs | Opus 4.5Anthropic | Claude Opus 4.6Anthropic |
|---|---|---|
Core specsRelease | 2025-11-25 | 2026-02-05 |
Context length | 200K | 1000K |
Max output | 65536 | 65536 |
MoE | No | No |
LicenseCode Open Source | Not provided | Not provided |
Weights Open Source | Not provided | Not provided |
Commercial use | 不开源 | 不开源 |
Modality supportText Input/Output | / | / |
Image Input/Output | / | / |
ResourcesPaper / report | Introducing Claude Opus 4.5 | Introducing Claude Opus 4.6 |
