Claude Opus 4.6 vs Claude Opus 4.7
Full profiles: Claude Opus 4.6 (Anthropic) and Claude Opus 4.7 (Anthropic): every benchmark, price and rank.
Updated · Data from 11 sources
As of 25 Oct 2026, Claude Opus 4.6 (high) and Claude Opus 4.7 are neck and neck on quality (62.7 vs 63.2, inside the ±3.5 margin of error) and cost about the same, so speed decides it: Claude Opus 4.7 streams 1.5× faster. Pick Claude Opus 4.6 for human preference (70 vs 68); pick Claude Opus 4.7 for maths (67 vs 64).
Claude Opus 4.6 benchmarks high Anthropic Claude Opus 4.7 benchmarks AnthropicOpen in the comparison toolWhich to pick for…
The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.
| Use case | Pick | Why | Confidence |
|---|---|---|---|
| Coding | Claude Opus 4.6 | 63 vs 61 category score; WeirdML 78.0% vs 76.4% | Medium |
| Maths | Claude Opus 4.7 | 67 vs 64 category score; Arena Maths 1491 vs 1517 | Medium |
| Reasoning | Either | 66 vs 65 category score; Arena Hard Prompts 1530 vs 1519 | Either |
| Images and other media | Either | 67 vs 67 category score; Arena Vision 1316 vs 1317 | Either |
| Chat quality (human preference) | Claude Opus 4.6 | 70 vs 68 on Arena votes | Medium |
| Lowest cost | Either | $10.00 vs $10.00 per 1M blended; agentic coding session $0.400 vs $0.400 | – |
| Fastest first answer | Either | 1.2 s (Claude Opus 4.6 (high)) vs 1.2 s (Claude Opus 4.7) to the first answer token | – |
| High volume on a budget | Neither flagship | A smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs | – |
Key differences
- Claude Opus 4.6 (high) and Claude Opus 4.7 are neck and neck on quality (62.7 vs 63.2, within the ±3.5 margin).
- Claude Opus 4.6 (high) is stronger in coding, human preference, reasoning.
- Claude Opus 4.7 is stronger in maths, multimodal, agents & tools, knowledge.
- They cost about the same ($10.00 per 1M blended).
- Claude Opus 4.7 streams 1.5× faster (60 vs 39 tokens per second).
At a glance
| Spec | Claude Opus 4.6 | Claude Opus 4.7 |
|---|---|---|
| Developer | Anthropic | Anthropic |
| Released | 5 Feb 2026 | 16 Apr 2026 |
| Context window (tokens) | 1,000,000 | 1,000,000 |
| Compared at | Claude Opus 4.6 (high) | Claude Opus 4.7 |
| Open weights | No | No |
| Price in / out / cached input per 1M | $5.00 / $25.00 / $0.500 | $5.00 / $25.00 / $0.500 |
Category by category
Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.
| Category | Claude Opus 4.6 (high) | Claude Opus 4.7 | Leader |
|---|---|---|---|
| Reasoning | 65.5 | 64.8 | level |
| Coding | 62.7 | 61.4 | Claude Opus 4.6 |
| Agents & tools | – | 65.4 | not comparable |
| Maths | 63.5 | 66.7 | Claude Opus 4.7 |
| Knowledge | – | 62.3 | not comparable |
| Human preference | 70.0 | 67.5 | Claude Opus 4.6 |
| Multimodal | 66.6 | 66.8 | level |
Where sources disagree
Across the 12 benchmarks both have, Claude Opus 4.6 leads 9, Claude Opus 4.7 leads 3. Independent sources split: Claude Opus 4.6 leads on Arena Text; Claude Opus 4.7 leads on AA Intelligence Index v4.3.2.
| Overall index | Claude Opus 4.6 (high) | Claude Opus 4.7 | Leader |
|---|---|---|---|
| BenchLeader Index | 62.7 | 63.2 | neck and neck |
| AA Intelligence Index v4.3.2 | 26.4† | 30.9 | Claude Opus 4.7 |
| Arena Text | 1503 | 1494 | Claude Opus 4.6 |
Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).
Price, speed and latency
List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page; suite cost is what Artificial Analysis paid to run its whole Intelligence Index.
| Measure | Claude Opus 4.6 (high) | Claude Opus 4.7 |
|---|---|---|
| Blended price per 1M tokens | $10.00 | $10.00 |
| Input / output per 1M | $5.00 / $25.00 | $5.00 / $25.00 |
| Cached input per 1M | $0.500 | $0.500 |
| Price per task: chat reply | $0.0095 | $0.0095 |
| Price per task: summarize a 30-page report | $0.075 | $0.075 |
| Price per task: code edit | $0.068 | $0.068 |
| Price per task: agentic coding session | $0.400 | $0.400 |
| Price per task: structured extraction | $0.015 | $0.015 |
| Output speed | 39 tok/s | 60 tok/s |
| Time to first answer | 1.2 s | 1.2 s |
Each model is compared at its best-ranked reasoning-effort setting. Claude Opus 4.6 (high) scores 62.7 and takes 1 s to its first answer token; Claude Opus 4.6 (thinking) scores 62.5 in 1 s, Claude Opus 4.6 scores 62.5 in 1 s. Claude Opus 4.7 scores 63.2 and takes 1 s to its first answer token; Claude Opus 4.7 (max) scores 62.4 in 23 s, Claude Opus 4.7 (high) scores 61.0 in 1 s. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.
Every benchmark
132 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.
| Metric | Publisher | Claude Opus 4.6 (high) | Claude Opus 4.7 |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 62.7 | 63.2 |
| Coding score | BenchLeader | 62.7 | 61.4 |
| Human preference score | BenchLeader | 70.0 | 67.5 |
| Maths score | BenchLeader | 63.5 | 66.7 |
| Multimodal score | BenchLeader | 66.6 | 66.8 |
| Reasoning score | BenchLeader | 65.5 | 64.8 |
| Agents & tools score | BenchLeader | – | 65.4 |
| Knowledge score | BenchLeader | – | 62.3 |
| Blended price $/M | BenchLeader | $10.00 | $10.00 |
| Output speed | BenchLeader | 39 tok/s | 60 tok/s |
| Time to first answer | BenchLeader | 1.2 s | 1.2 s |
| Context window | BenchLeader | 1,000,000 tokens | 1,000,000 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 90.5%† | 86.4%† |
| Humanity's Last Exam | Scale AI / CAIS | 19.0%† | 36.2% |
| SimpleBench | SimpleBench | 67.6%† | 61.7% |
| Arena Hard Prompts | Arena (LMArena) | 1530 | 1519 |
| LiveBench 1.0 Reasoning | LiveBench | 88.7% | 87.2%† |
| GPQA Diamond (AA) | Artificial Analysis | 89.6%† | 91.4%† |
| Humanity's Last Exam (AA) | Artificial Analysis | 39.9%† | 42.3%† |
| GPQA Diamond (Vals) | Vals AI | 89.7%† | 90.2% |
| Kagi LLM Benchmark | Kagi LLM Benchmark | 72.4%† | 73.3% |
| ARC-AGI-1 | ARC Prize | 94.0% | – |
| ARC-AGI-2 | ARC Prize | 69.2% | – |
| ARC-AGI-3 | ARC Prize | 0.5%† | 0.2%† |
| CritPt | Artificial Analysis | 12.6%† | 12.0%† |
| Chess Puzzles | Epoch AI Benchmarking Hub | 21.0%† | 41.0%† |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 15.0%† | 13.0% |
| CL-bench | Epoch AI Benchmarking Hub | 20.7%† | – |
| CL-bench Life | Epoch AI Benchmarking Hub | 17.0% | – |
| LMCA | Epoch AI Benchmarking Hub | 55.8%† | 52.2%† |
| DTBench | Epoch AI Benchmarking Hub | 91.2%† | 94.7%† |
| ForecastBench | ForecastBench | 63.4%† | 63.1% |
| Coding | |||
| SWE-bench Verified (Epoch) | Epoch AI Benchmarking Hub | 78.7%† | 83.5%† |
| SciCode | SciCode | – | 54.5%† |
| WeirdML | WeirdML | 78.0% | 76.4% |
| FrontierCode | Cognition | 26.6%† | 38.5% |
| GSO-Bench | GSO-Bench | 41.2% | 44.1% |
Show all 132 benchmarks
| Metric | Publisher | Claude Opus 4.6 (high) | Claude Opus 4.7 |
|---|---|---|---|
| Coding | |||
| Arena Coding | Arena (LMArena) | 1547 | 1546 |
| Arena WebDev | Arena (LMArena) | 1546 | 1555 |
| LiveBench 1.0 Coding | LiveBench | 78.2% | 82.1%† |
| LiveCodeBench | Vals AI | 84.7%† | 85.1% |
| SWE-bench (Vals) | Vals AI | 78.2%† | 82.0% |
| SWE-Bench Pro | Scale AI SEAL | 51.9%† | – |
| SWE-bench Verified (bash only) | SWE-bench | 75.6%† | – |
| SWE-bench Verified (any scaffold) | SWE-bench | 75.6%† | – |
| Code Migration | Vals AI | – | 43.9% |
| ProgramBench | Vals AI | – | 0.0% |
| Vibe Code Bench v1.1 | Vals AI | 57.6%† | 71.0% |
| SWE-Bench Pro (private) | Scale AI SEAL | 51.9%† | – |
| SWE Atlas: Codebase QnA | Scale AI SEAL | 33.3%† | 40.3% |
| SWE Atlas: Refactoring | Scale AI SEAL | 35.6%† | 48.6% |
| SWE Atlas: Test Writing | Scale AI SEAL | 36.7%† | 38.5% |
| MirrorCode | Epoch AI Benchmarking Hub | – | 31.1%† |
| ALE-Bench | Epoch AI Benchmarking Hub | 996.5† | 1323.0 |
| AlgoTune | AlgoTune | 1.5 | – |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | – | 83.2%† |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 79.8%† | 80.2% |
| OSWorld-Verified 2.0 | OSWorld | – | 18.2%† |
| Cybench | Cybench | 93.0%† | – |
| Remote Labor Index | Scale AI / CAIS | 4.2%† | – |
| APEX-Agents | Mercor | 46.3%† | 49.2%† |
| LiveBench Agentic Coding | LiveBench | 49.0% | 50.7%† |
| Terminal-Bench Hard | Artificial Analysis | 46.2%† | 51.5%† |
| τ²-Bench Telecom (AA) | Artificial Analysis | 92.1%† | 88.6%† |
| Terminal-Bench 2.1 (Vals) | Vals AI | – | 68.5% |
| MCP Atlas | Scale AI SEAL | 76.8%† | 79.1%† |
| HiL-Bench | Scale AI SEAL | 38.3%† | 41.7% |
| τ³-bench Banking | τ²-bench | 27.3%† | 40.2%† |
| τ³-Banking (AA) | Artificial Analysis | – | 34.6%† |
| ITBench SRE (AA) | Artificial Analysis | – | 46.7%† |
| Analyst Agent (AA) | Artificial Analysis | – | 43.8%† |
| APEX-Agents (AA) | Artificial Analysis | 33.0%† | – |
| Finance Agent v2 | Vals AI | – | 51.5% |
| Harvey's Legal Agent Benchmark v1 | Vals AI | – | 6.7% |
| Legal Research Bench | Vals AI | – | 38.5% |
| Tax Agent Bench | Vals AI | – | 23.7% |
| Terminal-Bench 2.0 (Vals) | Vals AI | 58.4%† | 68.5% |
| Arena Search | Arena (LMArena) | – | 1210 |
| DeepResearch Bench | DeepResearch Bench | 55.3% | – |
| PostTrainBench | Epoch AI Benchmarking Hub | – | 28.6%† |
| ExploitBench | Epoch AI Benchmarking Hub | – | 26.5% |
| METR Time Horizons | METR | 78.9%† | – |
| GBAEval | Epoch AI Benchmarking Hub | 44.1%† | 43.8% |
| Vending-Bench 2 | Andon Labs | 8017.6† | 10936.8 |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 66.0%† | 70.2%† |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 26.8%† | 31.7%† |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 94.4%† | 86.7%† |
| ProofBench | Vals AI | 50.0%† | 54.0%† |
| LiveBench Mathematics | LiveBench | 89.3% | 92.8%† |
| AIME (Vals) | Vals AI | 95.6%† | 96.3% |
| AIME 2026 | MathArena | 96.7% | 95.8%† |
| HMMT February 2026 | MathArena | 96.2% | 93.9%† |
| USAMO 2026 | MathArena | 47.0% | – |
| MathArena Apex | MathArena | 34.5% | 40.6%† |
| Arena Maths | Arena (LMArena) | 1517 | 1491 |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 47.0%† | 51.7%† |
| LiveBench 1.0 Data Analysis | LiveBench | 69.9% | 78.3%† |
| AA-Omniscience | Artificial Analysis | 13.7† | 27.3† |
| MMLU-Pro | Vals AI | 89.1%† | 89.9% |
| LegalBench | Vals AI | 85.3%† | 85.3% |
| CorpFin | Vals AI | 67.0%† | 66.1% |
| TaxEval | Vals AI | 76.0%† | 75.3% |
| MedQA | Vals AI | 95.4%† | – |
| PRBench Finance | Scale AI SEAL | 53.3%† | – |
| PRBench Legal | Scale AI SEAL | 52.3%† | – |
| MultiNRC | Scale AI SEAL | 57.1%† | – |
| CaseLaw v2 | Vals AI | 62.1%† | 68.4% |
| Excel Modeling Benchmark | Vals AI | – | 63.7% |
| MedCode | Vals AI | 48.2%† | 54.9% |
| MedScribe | Vals AI | 86.7%† | 83.0% |
| DrugDiscoveryBench | Scale AI SEAL | 27.7%† | – |
| EBR-bench | Epoch AI Benchmarking Hub | 12.7%† | 19.1%† |
| GDP.pdf | Epoch AI Benchmarking Hub | – | 21.0%† |
| AA-Omniscience: accuracy | Artificial Analysis | 47.0%† | 48.9%† |
| AA-Omniscience: non-hallucination | Artificial Analysis | 37.2%† | 57.7%† |
| Global-MMLU-Lite (AA) | Artificial Analysis | 92.2%† | – |
| Instruction following | |||
| LiveBench Language | LiveBench | 83.3% | 77.9%† |
| LiveBench Instruction Following | LiveBench | 63.3% | 66.7%† |
| IFBench | Artificial Analysis | 53.1%† | 58.6%† |
| MultiChallenge | Scale AI SEAL | 37.1%† | – |
| TutorBench | Scale AI SEAL | 53.7%† | – |
| Arena Instruction Following | Arena (LMArena) | 1511 | 1493 |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1503 | 1494 |
| EQ-Bench 4 | EQ-Bench | 1213† | 1294 |
| Arena Creative Writing | Arena (LMArena) | 1500 | 1482 |
| Arena Multi-turn | Arena (LMArena) | 1516 | 1514 |
| Arena Longer Queries | Arena (LMArena) | 1521 | 1508 |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1316 | 1317 |
| MMMU-Pro | Artificial Analysis | 75.4%† | 78.8%† |
| VISTA | Scale AI SEAL | 46.1%† | – |
| MMMU-Pro (Vals) | Vals AI | 83.9%† | 85.5% |
| MortgageTax | Vals AI | 68.5%† | 70.3% |
| SAGE | Vals AI | 51.6%† | 56.1% |
| VTB | Scale AI SEAL | 27.5%† | – |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | – | 24.5% |
| Long context | |||
| AA-LCR | Artificial Analysis | 78.0%† | 78.7%† |
| Arena Document | Arena (LMArena) | 1495 | 1494 |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 155.7† | 156.8 |
| LiveBench 1.0 | LiveBench | 74.5% | 76.5%† |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 31.9† | 40.7† |
| Vals RSI Index | Vals AI | – | 19.1%† |
| Vals Multimodal Index | Vals AI | – | 67.4% |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | 13.0%† | – |
| MASK | Scale AI SEAL | 85.4%† | – |
Data as of 2026-10-25. Best configuration of each model; every score links to its source on the model pages.
What changed
- Claude Opus 4.7: 1 score revision.
- Claude Opus 4.7: 1 new result (Harvey's Legal Agent Benchmark v1).
- Claude Opus 4.7: 1 score revision.
- Claude Opus 4.6 (high): BenchLeader Index 60.4 → 62.9.
- Claude Opus 4.7: 1 score revision.
- Claude Opus 4.6 (high): 1 new result (USAMO 2026).
- Claude Opus 4.7: 1 score revision.
- Claude Opus 4.7: BenchLeader Index 64.0 → 63.5.
- Claude Opus 4.7: 1 new result (SWE Atlas: Codebase QnA).
- Claude Opus 4.7: BenchLeader Index 64.5 → 64.0.
Related comparisons
- Claude Opus 4.6 vs Claude Opus 5.5
- Claude Opus 4.6 vs Gemini 4 Argon
- Claude Opus 4.6 vs Muse Wonder Full
- Claude Opus 4.7 vs Claude Opus 5.5
- Claude Opus 4.7 vs Gemini 4 Argon
- Claude Opus 4.7 vs Muse Wonder Full
- Claude Opus 4.7 vs Claude Fable 5.1
- Claude Opus 4.6 vs GPT-6 Astra
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
Claude Opus 4.6 vs Claude Opus 4.7: questions
- Is Claude Opus 4.6 better than Claude Opus 4.7?
- Claude Opus 4.6 (high) and Claude Opus 4.7 are neck and neck on quality (62.7 vs 63.2, within the ±3.5 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 2026-10-25, so check the category scores for your use.
- Which is cheaper, Claude Opus 4.6 or Claude Opus 4.7, and will that change?
- They cost the same: $10.00 per million tokens blended. No price change has been announced for either.
- Is Claude Opus 4.6 better than Claude Opus 4.7 for coding?
- Claude Opus 4.6 scores higher in coding (63 vs 61 on the category index, where 50 is average). On Arena WebDev, the top shared coding benchmark, it is 1546 for Claude Opus 4.6 against 1555 for Claude Opus 4.7.
- Which is better for long documents?
- Both accept 1,000,000 tokens of context.
- Which has lower latency?
- Claude Opus 4.6 (high) answers first: 1.2 s against 1.2 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
- Which is faster, Claude Opus 4.6 or Claude Opus 4.7?
- Claude Opus 4.7 streams faster: 60 against 39 output tokens per second.
- Do independent benchmarks agree on Claude Opus 4.6 vs Claude Opus 4.7?
- Across the 12 benchmarks both have, Claude Opus 4.6 leads 9, Claude Opus 4.7 leads 3. Independent sources split: Claude Opus 4.6 leads on Arena Text; Claude Opus 4.7 leads on AA Intelligence Index v4.3.2.
- When was Claude Opus 4.7 released?
- Claude Opus 4.7 was released on 16 Apr 2026.