Claude Opus 4.6 vs Claude Opus 4.7
New ComparisonScores are fetched live from the current database on every visit — nothing is cached.
Winner
Claude Opus 4.7 wins this comparison with an overall score of 81 vs 66 for Claude Opus 4.6, leading 5 of 7 shared benchmarks.
Based on 7 of 58 benchmarks scored by all 2 selected models.
Benchmark Scores Comparison
Claude Opus 4.6
Score: 15.0
Claude Opus 4.7
Score: 15.0
Coverage matrix
2 models × 58 benchmarks — hover a benchmark name for its category
| Benchmark | Claude Opus 4.6 Anthropic | Claude Opus 4.7 Anthropic |
|---|---|---|
| SVG-Bench | — | 62.3 |
| GPQA Diamond | 91.3 | — |
| LiveCodeBench v6 | 88.8 | — |
| SWE-bench Pro | 53.4 | 64.3 |
| MathVision | 71.2 | — |
| AIME 26 | 96.7 | — |
| LiveSQLBench | — | 41 |
| AgentWorldBench Terminal | 57.51 | — |
| Toolathlon Verified | 47.2 | — |
| OJBench | 60.3 | — |
| AgentWorldBench Overall | 57.8 | — |
| GDPVal | — | 79.8 |
| OfficeQA Pro | — | 43.6 |
| DeepSearch QA | 91.3 | — |
| SWE Atlas - TW | — | 38.2 |
| CL-bench | — | 22.9 |
| LOCA-Bench (256k) | — | 57 |
| KernelBench Hard | — | 30.7 |
| AgentWorldBench Web | 51.42 | — |
| SWE Atlas - QnA | — | 45.2 |
| IMO 2025 | — | 17.4 |
| MMMU-Pro | 73.9 | 77 |
| PostTrainBench | — | 42.4 |
| BabyVision | 14.8 | — |
| Claw-Eval Avg | 82.4 | 71.6 |
| VideoMMMU | — | 83 |
| Apex-Agents | 33 | 37.2 |
| BankerToolBench | — | 81.3 |
| OSWorld-Verified | 72.7 | 82.8 |
| AgentWorldBench Search | 29.3 | — |
| HLE (with tools) | 53 | — |
| CharXiv (RQ) | 69.1 | — |
| BrowseComp | 83.7 | 79.3 |
| NL2Repo-Bench | — | 56.3 |
| DRACO | — | 77.7 |
| MCPMark | 56.7 | — |
| SWE-bench Multilingual | 77.8 | — |
| PaperBench | — | 58.5 |
| YC-Bench | — | 2200000 |
| SWE-bench Verified | 80.8 | 87.6 |
| IMOAnswerBench | 75.3 | — |
| Humanity's Last Exam | 40 | — |
| Terminal-Bench 2.0 | 65.4 | — |
| SpreadSheetBench-v1 | — | 88.5 |
| VIBE-V2 | — | 55.9 |
| OmniDocBench 1.5 | — | 89.3 |
| HMMT Feb 26 | 96.2 | — |
| AgentWorldBench OS | 70.2 | — |
| AgentWorldBench MCP | 69.9 | — |
| Terminal Bench 2.1 | — | 66.1 |
| AgentWorldBench SWE | 64.55 | — |
| SciCode (subtask) | 51.9 | — |
| USAMO 2026 | — | 52.8 |
| Claw-Eval Pass^3 | 70.4 | — |
| MCP-Atlas | — | 77 |
| V-Star | 86.4 | — |
| AgentWorldBench Android | 61.74 | — |
| SWE-efficiency | — | 42.2 |
| Covered | 34/58 | 31/58 |
Scorecard
Overall 0–100 score — category strengths (50%), shared-benchmark wins (30%), coverage (20%)
| Model | Overall score | Shared benchmark wins | Coverage | Strongest at | Weakest at |
|---|---|---|---|---|---|
| Claude Opus 4.7 Winner | 80.8 | 5 of 7 | 31 of 58 (53.0%) | Vision & Language | Agentic Coding |
| Claude Opus 4.6 | 65.6 | 2 of 7 | 34 of 58 (59.0%) | Vision & Language | General Agents |
Category strengths
Where each model stands, per benchmark category — score relative to the best model, adjusted for benchmark coverage. This also ranks models when they share few benchmarks.
| Category | Claude Opus 4.7 | Claude Opus 4.6 |
|---|---|---|
| General Agents 21 benchmarks | 98.2 | 88.4 |
| Agentic Coding 18 benchmarks | 95.5 | 91.8 |
| Vision & Language 4 benchmarks | 100 | 93.9 |
Fully comparable benchmarks
Scored by all 2 selected models (7)
| Benchmark | Category | Claude Opus 4.6 | Claude Opus 4.7 |
|---|---|---|---|
| SWE-bench Pro | coding_agent | 53.4 | 64.3 |
| MMMU-Pro | vision_language | 73.9 | 77 |
| Claw-Eval Avg | coding_agent | 82.4 | 71.6 |
| Apex-Agents | general_agent | 33 | 37.2 |
| OSWorld-Verified | general_agent | 72.7 | 82.8 |
| BrowseComp | general_agent | 83.7 | 79.3 |
| SWE-bench Verified | coding_agent | 80.8 | 87.6 |
Partial coverage benchmarks
Scored by at least one but not all selected models (51) — missing scores show as —
| Benchmark | Category | Claude Opus 4.6 | Claude Opus 4.7 |
|---|---|---|---|
| SVG-Bench 1/2 models | coding_agent | — | 62.3 |
| GPQA Diamond 1/2 models | stem_reasoning | 91.3 | — |
| LiveCodeBench v6 1/2 models | stem_reasoning | 88.8 | — |
| MathVision 1/2 models | vision_language | 71.2 | — |
| AIME 26 1/2 models | stem_reasoning | 96.7 | — |
| LiveSQLBench 1/2 models | coding_agent | — | 41 |
| AgentWorldBench Terminal 1/2 models | general_agent | 57.51 | — |
| Toolathlon Verified 1/2 models | general_agent | 47.2 | — |
| OJBench 1/2 models | stem_reasoning | 60.3 | — |
| AgentWorldBench Overall 1/2 models | general_agent | 57.8 | — |
| GDPVal 1/2 models | general_agent | — | 79.8 |
| OfficeQA Pro 1/2 models | general_agent | — | 43.6 |
| DeepSearch QA 1/2 models | general_agent | 91.3 | — |
| SWE Atlas - TW 1/2 models | coding_agent | — | 38.2 |
| CL-bench 1/2 models | coding_agent | — | 22.9 |
| LOCA-Bench (256k) 1/2 models | long_context | — | 57 |
| KernelBench Hard 1/2 models | coding_agent | — | 30.7 |
| AgentWorldBench Web 1/2 models | general_agent | 51.42 | — |
| SWE Atlas - QnA 1/2 models | coding_agent | — | 45.2 |
| IMO 2025 1/2 models | stem_reasoning | — | 17.4 |
| PostTrainBench 1/2 models | coding_agent | — | 42.4 |
| BabyVision 1/2 models | vision_language | 14.8 | — |
| VideoMMMU 1/2 models | video_understanding | — | 83 |
| BankerToolBench 1/2 models | general_agent | — | 81.3 |
| AgentWorldBench Search 1/2 models | general_agent | 29.3 | — |
| HLE (with tools) 1/2 models | stem_reasoning | 53 | — |
| CharXiv (RQ) 1/2 models | document_understanding | 69.1 | — |
| NL2Repo-Bench 1/2 models | coding_agent | — | 56.3 |
| DRACO 1/2 models | general_agent | — | 77.7 |
| MCPMark 1/2 models | general_agent | 56.7 | — |
| SWE-bench Multilingual 1/2 models | coding_agent | 77.8 | — |
| PaperBench 1/2 models | coding_agent | — | 58.5 |
| YC-Bench 1/2 models | general_agent | — | 2200000 |
| IMOAnswerBench 1/2 models | stem_reasoning | 75.3 | — |
| Humanity's Last Exam 1/2 models | stem_reasoning | 40 | — |
| Terminal-Bench 2.0 1/2 models | coding_agent | 65.4 | — |
| SpreadSheetBench-v1 1/2 models | general_agent | — | 88.5 |
| VIBE-V2 1/2 models | coding_agent | — | 55.9 |
| OmniDocBench 1.5 1/2 models | document_understanding | — | 89.3 |
| HMMT Feb 26 1/2 models | stem_reasoning | 96.2 | — |
| AgentWorldBench OS 1/2 models | general_agent | 70.2 | — |
| AgentWorldBench MCP 1/2 models | general_agent | 69.9 | — |
| Terminal Bench 2.1 1/2 models | coding_agent | — | 66.1 |
| AgentWorldBench SWE 1/2 models | general_agent | 64.55 | — |
| SciCode (subtask) 1/2 models | stem_reasoning | 51.9 | — |
| USAMO 2026 1/2 models | stem_reasoning | — | 52.8 |
| Claw-Eval Pass^3 1/2 models | coding_agent | 70.4 | — |
| MCP-Atlas 1/2 models | general_agent | — | 77 |
| V-Star 1/2 models | vision_language | 86.4 | — |
| AgentWorldBench Android 1/2 models | general_agent | 61.74 | — |
| SWE-efficiency 1/2 models | coding_agent | — | 42.2 |
Comparison FAQ
Which model is better, Claude Opus 4.7 or Claude Opus 4.6?
Claude Opus 4.7 wins this comparison with an overall score of 81 vs 66 for Claude Opus 4.6, leading 5 of 7 shared benchmarks.
Which model is better at General Agents, Claude Opus 4.7 or Claude Opus 4.6?
Claude Opus 4.7 leads the General Agents category (strength score 98.2/100 across 3 comparable benchmarks).
How reliable is this comparison verdict?
Based on 7 of 58 benchmarks scored by all 2 selected models.