Claude Opus 4.5 vs Claude Sonnet 5
Full profiles: Claude Opus 4.5 (Anthropic) and Claude Sonnet 5 (Anthropic): every benchmark, price and rank.
Updated · Data from 11 sources
As of 16 Nov 2026, Claude Opus 4.5 and Claude Sonnet 5 are neck and neck on quality (59.7 vs 59.0 on the BenchLeader Index, inside the ±4.1 margin), and Claude Sonnet 5 is 2.5× cheaper at $4.00 vs $10.00 per 1M tokens blended. Claude Opus 4.5 leads in every category we measure both in. Claude Sonnet 5 streams ~2× faster.
Key facts (as of 16 Nov 2026)
- Claude Opus 4.5 scores 59.7 on the BenchLeader Index and Claude Sonnet 5 scores 59.0 (BenchLeader, 16 Nov 2026).
- Claude Opus 4.5 costs $10.00 per 1M tokens blended (list price, 16 Nov 2026).
- Claude Sonnet 5 costs $4.00 per 1M tokens blended (list price, 16 Nov 2026).
- Claude Sonnet 5 streams ~2× faster than Claude Opus 4.5 (Artificial Analysis, 7-day median, 16 Nov 2026).
- Of the 11 benchmarks both models have, Claude Opus 4.5 leads 8 (BenchLeader, 16 Nov 2026).
- On Arena WebDev, Claude Opus 4.5 is rated 1493 and Claude Sonnet 5 1539 (LMArena, as published by 16 Nov 2026).
Which 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 | Either | 61 vs 60 category score; Arena WebDev 1493 vs 1539 | Either |
| Agents & tool use | Either | 53 vs 53 category score | Either |
| Maths | Either | 65 vs 65 category score; Arena Maths 1472 vs 1472 | Either |
| Reasoning | Claude Opus 4.5 | 67 vs 65 category score; Arena Hard Prompts 1501 vs 1490 | Medium |
| Chat quality (human preference) | Claude Opus 4.5 | 66 vs 65 on Arena votes | Medium |
| Lowest cost | Claude Sonnet 5 | $4.00 vs $10.00 per 1M blended; agentic coding session $0.160 vs $0.400 | – |
| Fastest first answer | Claude Opus 4.5 | 1.1 s (Claude Opus 4.5 (high)) vs 15.8 s (Claude Sonnet 5 (high)) 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 4.5 (high) and Claude Sonnet 5 (high) are neck and neck on quality (59.7 vs 59.0, within the ±4.1 margin).
- Claude Opus 4.5 (high) is stronger in coding, human preference, reasoning.
- Claude Sonnet 5 (high) is stronger in agents & tools.
- Claude Sonnet 5 (high) is 2.5× cheaper ($4.00 vs $10.00 per 1M blended).
- Claude Sonnet 5 (high) streams 2.5× faster (99 vs 40 tokens per second).
At a glance
| Spec | Claude Opus 4.5 | Claude Sonnet 5 |
|---|---|---|
| Developer | Anthropic | Anthropic |
| Released | 24 Nov 2025 | 30 Jun 2026 |
| Context window (tokens) | 200,000 | 1,000,000 |
| Compared at | Claude Opus 4.5 (high) | Claude Sonnet 5 (high) |
| Open weights | No | No |
| Price in / out / cached input per 1M | $5.00 / $25.00 / $0.500 | $2.00 / $10.00 / $0.200 |
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 4.5 (high) | Claude Sonnet 5 (high) | Leader |
|---|---|---|---|
| Reasoning | 66.6 | 65.4 | Claude Opus 4.5 |
| Coding | 60.6 | 59.9 | level |
| Agents & tools | 52.7 | 52.9 | level |
| Maths | 65.3 | 65.3 | level |
| Knowledge | 60.0 | – | not comparable |
| Human preference | 66.2 | 64.6 | Claude Opus 4.5 |
| Multimodal | – | 60.9 | not comparable |
| Long context | – | 62.9 | not comparable |
On one chart
Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.
Not drawn, measured for fewer than two of these models: Knowledge, Instruction following, Long context, Multimodal.
Where sources disagree
Across the 11 benchmarks both have, Claude Opus 4.5 leads 8, Claude Sonnet 5 leads 3. Every overall index that rates both puts Claude Opus 4.5 ahead.
| Overall index | Claude Opus 4.5 (high) | Claude Sonnet 5 (high) | Leader |
|---|---|---|---|
| BenchLeader Index | 59.7 | 59.0 | neck and neck |
| Arena Text | 1474 | 1461 | Claude Opus 4.5 |
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. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.
| Measure | Claude Opus 4.5 (high) | Claude Sonnet 5 (high) |
|---|---|---|
| Blended price per 1M tokens | $10.00 | $4.00 |
| Input / output per 1M | $5.00 / $25.00 | $2.00 / $10.00 |
| Cached input per 1M | $0.500 | $0.200 |
| Price per task: chat reply | $0.0095 | $0.0038 |
| Price per task: summarize a 30-page report | $0.075 | $0.030 |
| Price per task: code edit | $0.068 | $0.027 |
| Price per task: agentic coding session | $0.400 | $0.160 |
| Price per task: structured extraction | $0.015 | $0.0060 |
| Agentic coding session, input mostly cached | $0.198 | $0.079 |
| Output speed | 40 tok/s | 99 tok/s |
| Time to first answer | 1.1 s | 15.8 s |
| Time to a complete answer, last measured 13 Nov 2026 | – | 40.7 s |
Each model is compared at its best-ranked reasoning-effort setting. Claude Opus 4.5 (high) scores 59.7 and takes 1 s to its first answer token; Claude Opus 4.5 (thinking) scores 59.3 in 14 s, Claude Opus 4.5 scores 57.1 in 1 s. Claude Sonnet 5 (high) scores 59.0 and takes 16 s to its first answer token; Claude Sonnet 5 (max) scores 56.1 in 158 s, Claude Sonnet 5 (xhigh) scores 55.4 in 32 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
162 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 4.5 (high) | Claude Sonnet 5 (high) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 59.7 | 59.0 |
| Agents & tools score | BenchLeader | 52.7 | 52.9 |
| Coding score | BenchLeader | 60.6 | 59.9 |
| Human preference score | BenchLeader | 66.2 | 64.6 |
| Knowledge score | BenchLeader | 60.0 | – |
| Maths score | BenchLeader | 65.3 | 65.3 |
| Reasoning score | BenchLeader | 66.6 | 65.4 |
| Long context score | BenchLeader | – | 62.9 |
| Multimodal score | BenchLeader | – | 60.9 |
| Blended price $/M | BenchLeader | $10.00 | $4.00 |
| Output speed | BenchLeader | 40 tok/s | 99 tok/s |
| Time to first answer | BenchLeader | 1.1 s | 15.8 s |
| Context window | BenchLeader | 200,000 tokens | 1,000,000 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 86.0%† | 80.3%† |
| Humanity's Last Exam | Scale AI / CAIS | 25.2%† | – |
| SimpleBench | SimpleBench | 62.0%† | 60.6%† |
| Arena Hard Prompts | Arena (LMArena) | 1501 | 1490 |
| LiveBench 1.0 Reasoning | LiveBench | 80.1% | 88.7%† |
| GPQA Diamond (AA) | Artificial Analysis | 86.6%† | 91.1%† |
| Humanity's Last Exam (AA) | Artificial Analysis | 30.1%† | 35.7% |
| GPQA Diamond (Vals) | Vals AI | 85.9% | 88.9%† |
| Kagi LLM Benchmark | Kagi LLM Benchmark | 70.7%† | – |
| ARC-AGI-1 | ARC Prize | 80.0%† | – |
| ARC-AGI-2 | ARC Prize | 37.6%† | – |
| CritPt | Artificial Analysis | 4.6%† | 15.1% |
| SciPredict | Scale AI SEAL | 23.1%† | – |
| Arena Expert | Arena (LMArena) | 1503 | 1509 |
| Chess Puzzles | Epoch AI Benchmarking Hub | 14.0%† | 45.0%† |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 22.0%† | 16.0%† |
| CL-bench | Epoch AI Benchmarking Hub | 21.1%† | – |
| LMCA | Epoch AI Benchmarking Hub | 44.5% | 50.0% |
| DTBench | Epoch AI Benchmarking Hub | 89.9% | 84.5% |
| ForecastBench | ForecastBench | 64.6%† | 63.1%† |
| BTF-3 | Epoch AI Benchmarking Hub | – | 14.2%† |
| Coding | |||
| SWE-bench Verified (Epoch) | Epoch AI Benchmarking Hub | 76.7%† | – |
| SciCode | SciCode | – | 54.3% |
| WeirdML | WeirdML | 63.7%† | 68.8% |
| FrontierCode | Cognition | – | 42.7%† |
Show all 162 benchmarks
| Metric | Publisher | Claude Opus 4.5 (high) | Claude Sonnet 5 (high) |
|---|---|---|---|
| Coding | |||
| GSO-Bench | GSO-Bench | 26.5%† | 37.3%† |
| Arena Coding | Arena (LMArena) | 1530 | 1520 |
| Arena WebDev | Arena (LMArena) | 1493 | 1539 |
| LiveBench 1.0 Coding | LiveBench | 79.7% | 80.7%† |
| LiveCodeBench | Vals AI | 83.7% | 82.4%† |
| IOI | Vals AI | – | 45.0%† |
| SWE-bench (Vals) | Vals AI | 76.4% | 79.6%† |
| SWE-Bench Pro | Scale AI SEAL | 45.9%† | – |
| SWE-bench Verified (bash only) | SWE-bench | 76.8% | – |
| SWE-bench Multilingual (bash only) | SWE-bench | 70.7%† | – |
| SWE-bench Verified (any scaffold) | SWE-bench | 79.2%† | – |
| Code Migration | Vals AI | – | 44.4%† |
| ProgramBench | Vals AI | – | 0.0%† |
| Vibe Code Bench 1-100 | Vals AI | – | 13.8%† |
| Vibe Code Bench v1.1 | Vals AI | 20.6% | 81.3%† |
| SWE-Bench Pro (private) | Scale AI SEAL | 23.4%† | – |
| SWE-Bench Pro V2 | Scale AI SEAL | – | 93.2%† |
| SWE-Bench Pro V2 HARD | Scale AI SEAL | – | 88.2%† |
| MirrorCode | Epoch AI Benchmarking Hub | – | 63.3%† |
| Surface Evolver Bench | Epoch AI Benchmarking Hub | – | 60.0%† |
| DeepSWE v1.1 | Epoch AI Benchmarking Hub | – | 48.2% |
| CursorBench | Cursor | – | 30.8% |
| ALE-Bench | Epoch AI Benchmarking Hub | 1025.4† | 1463.1 |
| AlgoTune | AlgoTune | 1.8† | – |
| WeirdML v3 | WeirdML | 6.5% | – |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | – | 5.0% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | – | 80.5%† |
| Startrise Frontend | Startrise | – | 66.6%† |
| GameBench 2 Overall | DuelLab GameBench 2 | 47.1† | 46.6† |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | – | 54.4† |
| Design Arena Website | Design Arena | 1239† | 1268† |
| FrontierCode 2.0 (AA) | Artificial Analysis | – | 17.8%† |
| LiveCodeBench (AA) | Artificial Analysis | 87.1%† | – |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 63.1%† | 12.1%† |
| GDPval | OpenAI | 45.5%† | – |
| Cybench | Cybench | 82.0%† | – |
| Remote Labor Index | Scale AI / CAIS | 3.8%† | – |
| APEX-Agents | Mercor | – | 54.5%† |
| Arena Agent | Arena (LMArena) | – | 2 |
| LiveBench Agentic Coding | LiveBench | 39.7% | 59.4%† |
| Terminal-Bench Hard | Artificial Analysis | 47.0%† | – |
| τ²-Bench Telecom (AA) | Artificial Analysis | 89.5%† | – |
| Terminal-Bench 2.1 (Vals) | Vals AI | – | 74.5%† |
| MCP Atlas | Scale AI SEAL | 69.8% | – |
| BFCL Overall | Berkeley Function Calling Leaderboard | 77.5%† | – |
| τ²-bench | τ²-bench | 85.3% | – |
| τ³-bench Banking | τ²-bench | 24.7% | – |
| GDPval-AA v2.1 | Artificial Analysis | – | 1262 |
| τ³-Banking (AA) | Artificial Analysis | – | 37.3%† |
| Analyst Agent (AA) | Artificial Analysis | – | 46.3%† |
| CyberBench | Vals AI | – | 61.9%† |
| Finance Agent v2 | Vals AI | – | 53.9%† |
| Harvey's Legal Agent Benchmark v1 | Vals AI | – | 5.0%† |
| Legal Research Bench | Vals AI | – | 41.8%† |
| Poker Agent | Vals AI | 1033.4%† | – |
| SkillsBench | Vals AI | – | 46.5%† |
| SREBench | Vals AI | – | 0.4%† |
| Tax Agent Bench | Vals AI | – | 29.4%† |
| Terminal-Bench 2.0 (Vals) | Vals AI | 58.4% | – |
| Terminal-Bench 4.0 (Vals) | Vals AI | – | 10.6%† |
| Terminal-Bench Science | Vals AI | – | 5.7%† |
| Arena Search | Arena (LMArena) | 1186† | 1191† |
| Arena Agent Task Completion | Arena (LMArena) | – | -0.6 |
| BALROG | BALROG | 43.0%† | – |
| DeepResearch Bench | DeepResearch Bench | 54.8% | – |
| METR Time Horizons | METR | 75.0%† | – |
| GBAEval | Epoch AI Benchmarking Hub | – | 65.3%† |
| Vending-Bench 2 | Andon Labs | 4967.1† | 6377.7† |
| AutomationBench | Artificial Analysis | – | 32.1% |
| GDP.pdf (AA) | Artificial Analysis | – | 9.4% |
| EnterpriseOps-Gym | Artificial Analysis | – | 44.7%† |
| AA-Briefcase v1.1 | Artificial Analysis | – | 1177 |
| Design Arena Agentic | Design Arena | 1156† | 1222† |
| DeepSearchQA | 24.0%† | – | |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 34.4%† | 65.6%† |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 4.9%† | 29.3%† |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 86.1%† | 80.0%† |
| ProofBench | Vals AI | 36.0%† | 77.0%† |
| LiveBench Mathematics | LiveBench | 90.4% | 92.9%† |
| AIME (Vals) | Vals AI | 76.9%† | – |
| MGSM | Vals AI | 94.8%† | – |
| Arena Maths | Arena (LMArena) | 1472 | 1472 |
| AIME 2025 (AA) | Artificial Analysis | 91.3%† | – |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 45.7%† | 33.7%† |
| LiveBench 1.0 Data Analysis | LiveBench | 74.4% | 71.7%† |
| MMLU-Pro | Vals AI | 87.3% | 87.5%† |
| LegalBench | Vals AI | 84.6% | 83.9%† |
| CorpFin | Vals AI | 65.1% | 67.0%† |
| TaxEval | Vals AI | 74.9% | 75.6%† |
| MedQA | Vals AI | 93.2%† | – |
| PRBench Finance | Scale AI SEAL | 46.2%† | – |
| PRBench Legal | Scale AI SEAL | 44.2%† | – |
| MultiNRC | Scale AI SEAL | 41.2%† | – |
| CaseLaw v2 | Vals AI | 62.6%† | – |
| Excel Modeling Benchmark | Vals AI | – | 66.3%† |
| MedCode | Vals AI | 49.2% | 47.5%† |
| MedScribe | Vals AI | 85.3% | 76.0%† |
| Public Benefits Bench v1.1 | Vals AI | – | 66.0%† |
| DrugDiscoveryBench | Scale AI SEAL | – | 50.0%† |
| EBR-bench | Epoch AI Benchmarking Hub | 14.3%† | – |
| Global-MMLU-Lite (AA) | Artificial Analysis | 91.3%† | – |
| FACTS Parametric | – | 20.5%† | |
| Instruction following | |||
| LiveBench Language | LiveBench | 81.3% | 75.0%† |
| LiveBench Instruction Following | LiveBench | 62.5% | 63.9%† |
| IFBench | Artificial Analysis | 58.0%† | – |
| MultiChallenge | Scale AI SEAL | 59.0%† | – |
| TutorBench | Scale AI SEAL | 49.8%† | – |
| Arena Instruction Following | Arena (LMArena) | 1483 | 1467 |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1474 | 1461 |
| EQ-Bench 4 | EQ-Bench | – | 1221† |
| EQ-Bench Creative Writing v3 | EQ-Bench | 1686† | 1798† |
| Arena Creative Writing | Arena (LMArena) | 1469 | 1434 |
| Arena Multi-turn | Arena (LMArena) | 1485 | 1473 |
| Arena Longer Queries | Arena (LMArena) | 1495 | 1481 |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | – | 1272 |
| MMMU-Pro | Artificial Analysis | 74.0%† | 77.3%† |
| VISTA | Scale AI SEAL | 45.3%† | – |
| MMMU-Pro (Vals) | Vals AI | 83.0% | 83.0%† |
| MortgageTax | Vals AI | 68.7% | 70.0%† |
| Humanity's Sixth Sense | Scale AI SEAL | – | 24.8%† |
| GeoBench | GeoBench | 75.0%† | 74.0% |
| VPCT | Epoch AI Benchmarking Hub | 40.0%† | – |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | – | 24.9%† |
| Furniture Assembly | Epoch AI Benchmarking Hub | 28.3%† | – |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 77.3%† | 76.7% |
| Arena Document | Arena (LMArena) | 1468† | 1469 |
| MLCR | Artificial Analysis | – | 4.4% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 151.9† | 156.9† |
| LiveBench 1.0 | LiveBench | 72.6% | 76.0%† |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 29.1† | 31.7 |
| AA Progress Index v1 | Artificial Analysis | 284.4† | 296.8 |
| Vals RSI Index | Vals AI | 14.8% | – |
| Vals Multimodal Index | Vals AI | – | 68.8%† |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | 13.6%† | – |
| MASK | Scale AI SEAL | 87.1%† | – |
| DistressBench | Scale AI SEAL | – | 68.0%† |
| Metadata | |||
| AA Openness Index | Artificial Analysis | 11.1† | – |
Data as of 2026-11-16. Best configuration of each model; every score links to its source on the model pages.
DuelLab GameBench 2, release gb2-20261111-7b598bb13de4 (https://benchmarks.duellab.org/releases/gb2-20261111-7b598bb13de4/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.
What changed
- Claude Opus 4.5 (high): 13 new results (LiveCodeBench, MMLU-Pro, LegalBench…).
- Claude Sonnet 5 (high): 2 new results (GDPval-AA v2.1, AA-Briefcase v1.1).
- Claude Sonnet 5 (high): BenchLeader Index 59.5 → 59.0.
- Claude Sonnet 5 (high): 1 new result (GeoBench).
- Claude Sonnet 5 (high): 1 score revision.
- Claude Sonnet 5 (high): 1 score revision.
- Claude Sonnet 5 (high): 1 new result (AA Progress Index v1).
- Claude Opus 4.5 (high): 1 new result (Vals RSI Index).
- Claude Sonnet 5 (high): BenchLeader Index 60.0 → 59.5.
- Claude Sonnet 5 (high): 1 new result (MLCR).
Related comparisons
- Claude Sonnet 5 vs Gemini 4 Carbon
- Claude Sonnet 5 vs Claude Opus 5.5
- Claude Sonnet 5 vs GPT-6 Astra
- Claude Opus 4.5 vs GPT-6.1 Sol
- Claude Opus 4.5 vs Gemini 4 Argon
- Claude Opus 4.5 vs Claude Sonnet 5.5
- Claude Opus 4.5 vs Muse Wonder Full
- Claude Opus 4.5 vs Muse Spark 1.3
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
Claude Opus 4.5 vs Claude Sonnet 5: questions
- Is Claude Opus 4.5 better than Claude Sonnet 5?
- Claude Opus 4.5 (high) and Claude Sonnet 5 (high) are neck and neck on quality (59.7 vs 59.0, within the ±4.1 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 16 Nov 2026, so check the category scores for your use.
- Claude Opus 4.5 vs Claude Sonnet 5: which is cheaper, and will that change?
- Claude Sonnet 5 is cheaper today: $4.00 against $10.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is Claude Opus 4.5 or Claude Sonnet 5 better for coding?
- They are level in coding (61 vs 60 on the category index). On Arena WebDev, the top shared coding benchmark, it is 1493 for Claude Opus 4.5 against 1539 for Claude Sonnet 5.
- Which is faster, Claude Opus 4.5 or Claude Sonnet 5?
- Claude Sonnet 5 streams ~2× faster (99 vs 40 output tokens per second, Artificial Analysis 7-day median).
- Is Claude Opus 4.5 or Claude Sonnet 5 better for agents and tool use?
- They are level in agentic tasks (53 vs 53 on the category index).
More questions (4)
- Which is better for long documents?
- Claude Sonnet 5 accepts more context: 1,000,000 against 200,000 tokens for Claude Opus 4.5.
- Which has lower latency?
- Claude Opus 4.5 answers first: 1.1 s against 15.8 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.
- Do independent benchmarks agree on Claude Opus 4.5 vs Claude Sonnet 5?
- Across the 11 benchmarks both have, Claude Opus 4.5 leads 8, Claude Sonnet 5 leads 3. Every overall index that rates both puts Claude Opus 4.5 ahead.
- When was Claude Sonnet 5 released?
- Claude Sonnet 5 was released on 30 Jun 2026.