Claude Sonnet 5 vs o3
Full profiles: Claude Sonnet 5 (Anthropic) and o3 (OpenAI): every benchmark, price and rank.
Updated · Data from 16 sources
As of 16 Nov 2026, Claude Sonnet 5 and o3 are neck and neck on quality (59.0 vs 60.4 on the BenchLeader Index, inside the ±4.1 margin), and they cost about the same at $4.00 vs $3.50 per 1M tokens blended. Choose Claude Sonnet 5 for multimodal (61 vs 53), reasoning (65 vs 61), and human preference (65 vs 61); choose o3 for fast first answers (5.2 s vs 15.8 s) and agents & tools (70 vs 53) and maths (71 vs 65).
Key facts (as of 16 Nov 2026)
- Claude Sonnet 5 scores 59.0 on the BenchLeader Index and o3 scores 60.4 (BenchLeader, 16 Nov 2026).
- Claude Sonnet 5 costs $4.00 per 1M tokens blended (list price, 16 Nov 2026).
- o3 costs $3.50 per 1M tokens blended (list price, 16 Nov 2026).
- Of the 13 benchmarks both models have, Claude Sonnet 5 leads 13 (BenchLeader, 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 | 60 vs 60 category score; Arena Coding 1520 vs 1460 | Either |
| Agents & tool use | o3 | 70 vs 53 category score | High |
| Maths | o3 | 71 vs 65 category score; Arena Maths 1448 vs 1472 | Medium |
| Reasoning | Claude Sonnet 5 | 65 vs 61 category score; Arena Hard Prompts 1490 vs 1441 | Medium |
| Long documents | Claude Sonnet 5 | 63 vs 62 category score; AA-LCR v1.1 76.7% vs 74.7% | Medium |
| Images and other media | Claude Sonnet 5 | 61 vs 53 category score; Arena Vision 1272 vs 1214 | High |
| Chat quality (human preference) | Claude Sonnet 5 | 65 vs 61 on Arena votes | Medium |
| Lowest cost | Either | $4.00 vs $3.50 per 1M blended; agentic coding session $0.160 vs $0.152 | – |
| Fastest first answer | o3 | 5.2 s (o3) 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 Sonnet 5 (high) and o3 are neck and neck on quality (59.0 vs 60.4, within the ±4.1 margin).
- Claude Sonnet 5 (high) is stronger in coding, human preference, long context, multimodal, reasoning.
- o3 is stronger in agents & tools, maths.
- They cost about the same ($3.50 per 1M blended).
- o3 streams 1.3× faster (131 vs 99 tokens per second).
At a glance
| Spec | Claude Sonnet 5 | o3 |
|---|---|---|
| Developer | Anthropic | OpenAI |
| Released | 30 Jun 2026 | 16 Apr 2025 |
| Context window (tokens) | 1,000,000 | 200,000 |
| Compared at | Claude Sonnet 5 (high) | o3 |
| Open weights | No | No |
| Price in / out / cached input per 1M | $2.00 / $10.00 / $0.200 | $2.00 / $8.00 / $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 Sonnet 5 (high) | o3 | Leader |
|---|---|---|---|
| Reasoning | 65.4 | 60.9 | Claude Sonnet 5 |
| Coding | 59.9 | 59.8 | level |
| Agents & tools | 52.9 | 70.3 | o3 |
| Maths | 65.3 | 70.6 | o3 |
| Knowledge | – | 56.2 | not comparable |
| Instruction following | – | 65.2 | not comparable |
| Human preference | 64.6 | 61.0 | Claude Sonnet 5 |
| Multimodal | 60.9 | 52.8 | Claude Sonnet 5 |
| Long context | 62.9 | 61.9 | Claude Sonnet 5 |
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.
Where sources disagree
Across the 13 benchmarks both have, Claude Sonnet 5 leads 13, o3 leads 0. Every overall index that rates both puts Claude Sonnet 5 ahead.
| Overall index | Claude Sonnet 5 (high) | o3 | Leader |
|---|---|---|---|
| BenchLeader Index | 59.0 | 60.4 | neck and neck |
| Arena Text | 1461 | 1432 | Claude Sonnet 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 Sonnet 5 (high) | o3 |
|---|---|---|
| Blended price per 1M tokens | $4.00 | $3.50 |
| Input / output per 1M | $2.00 / $10.00 | $2.00 / $8.00 |
| Cached input per 1M | $0.200 | $0.500 |
| Price per task: chat reply | $0.0038 | $0.0032 |
| Price per task: summarize a 30-page report | $0.030 | $0.029 |
| Price per task: code edit | $0.027 | $0.024 |
| Price per task: agentic coding session | $0.160 | $0.152 |
| Price per task: structured extraction | $0.0060 | $0.0056 |
| Agentic coding session, input mostly cached | $0.079 | $0.085 |
| Output speed | 99 tok/s | 131 tok/s |
| Time to first answer | 15.8 s | 5.2 s |
| Time to a complete answer, last measured 13 Nov 2026 | 40.7 s | 17.3 s |
Each model is compared at its best-ranked reasoning-effort setting. 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. o3 scores 60.4 and takes 5 s to its first answer token; o3 (high) scores 55.1 in 5 s, o3 (medium) scores 53.3 in 5 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
159 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 Sonnet 5 (high) | o3 |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 59.0 | 60.4 |
| Agents & tools score | BenchLeader | 52.9 | 70.3 |
| Coding score | BenchLeader | 59.9 | 59.8 |
| Human preference score | BenchLeader | 64.6 | 61.0 |
| Long context score | BenchLeader | 62.9 | 61.9 |
| Maths score | BenchLeader | 65.3 | 70.6 |
| Multimodal score | BenchLeader | 60.9 | 52.8 |
| Reasoning score | BenchLeader | 65.4 | 60.9 |
| Instruction following score | BenchLeader | – | 65.2 |
| Knowledge score | BenchLeader | – | 56.2 |
| Blended price $/M | BenchLeader | $4.00 | $3.50 |
| Output speed | BenchLeader | 99 tok/s | 131 tok/s |
| Time to first answer | BenchLeader | 15.8 s | 5.2 s |
| Context window | BenchLeader | 1,000,000 tokens | 200,000 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 80.3%† | 81.8%† |
| Humanity's Last Exam | Scale AI / CAIS | – | 20.3%† |
| SimpleBench | SimpleBench | 60.6%† | 53.1%† |
| Arena Hard Prompts | Arena (LMArena) | 1490 | 1441 |
| LiveBench 1.0 Reasoning | LiveBench | 88.7%† | – |
| GPQA Diamond (AA) | Artificial Analysis | 91.1%† | 82.7% |
| Humanity's Last Exam (AA) | Artificial Analysis | 35.7% | 20.1% |
| GPQA Diamond (Vals) | Vals AI | 88.9%† | 84.1%† |
| Kagi LLM Benchmark | Kagi LLM Benchmark | – | 67.6% |
| GPQA Diamond (HELM) | HELM Capabilities | – | 75.3% |
| ARC-AGI-1 | ARC Prize | – | 60.8%† |
| ARC-AGI-2 | ARC Prize | – | 6.5%† |
| CritPt | Artificial Analysis | 15.1% | 1.1% |
| SciPredict | Scale AI SEAL | – | 17.9% |
| Arena Expert | Arena (LMArena) | 1509 | 1448 |
| Chess Puzzles | Epoch AI Benchmarking Hub | 45.0%† | 36.0%† |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 16.0%† | 29.0%† |
| CL-bench | Epoch AI Benchmarking Hub | – | 17.8%† |
| LMCA | Epoch AI Benchmarking Hub | 50.0% | 39.7%† |
| DTBench | Epoch AI Benchmarking Hub | 84.5% | 84.8%† |
| ForecastBench | ForecastBench | 63.1%† | 64.3% |
| BTF-3 | Epoch AI Benchmarking Hub | 14.2%† | – |
| Coding | |||
| SWE-bench Verified (Epoch) | Epoch AI Benchmarking Hub | – | 62.3%† |
| SciCode | SciCode | 54.3% | – |
| WeirdML | WeirdML | 68.8% | 52.4%† |
Show all 159 benchmarks
| Metric | Publisher | Claude Sonnet 5 (high) | o3 |
|---|---|---|---|
| Coding | |||
| FrontierCode | Cognition | 42.7%† | – |
| GSO-Bench | GSO-Bench | 37.3%† | 8.8%† |
| Arena Coding | Arena (LMArena) | 1520 | 1460 |
| Arena WebDev | Arena (LMArena) | 1539 | – |
| LiveBench 1.0 Coding | LiveBench | 80.7%† | – |
| LiveCodeBench | Vals AI | 82.4%† | 83.9%† |
| IOI | Vals AI | 45.0%† | – |
| SWE-bench (Vals) | Vals AI | 79.6%† | – |
| Aider Polyglot | Aider polyglot leaderboard | – | 76.9% |
| SWE-bench Verified (bash only) | SWE-bench | – | 58.4% |
| SWE-bench Verified (any scaffold) | SWE-bench | – | 58.4% |
| 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 | 81.3%† | – |
| 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%† | – |
| CadEval | Epoch AI Benchmarking Hub | – | 74.0%† |
| 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 | 1463.1 | 933.5† |
| 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 | 46.6† | 34.1 |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | 54.4† | 30.4† |
| Design Arena Website | Design Arena | 1268† | 1028 |
| FrontierCode 2.0 (AA) | Artificial Analysis | 17.8%† | – |
| LiveCodeBench (AA) | Artificial Analysis | – | 80.8% |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 12.1%† | – |
| GDPval | OpenAI | – | 30.8%† |
| APEX-Agents | Mercor | 54.5%† | – |
| Arena Agent | Arena (LMArena) | 2 | – |
| LiveBench Agentic Coding | LiveBench | 59.4%† | – |
| Terminal-Bench Hard | Artificial Analysis | – | 37.1% |
| τ²-Bench Telecom (AA) | Artificial Analysis | – | 80.7% |
| Terminal-Bench 2.1 (Vals) | Vals AI | 74.5%† | – |
| BFCL Overall | Berkeley Function Calling Leaderboard | – | 63.0% |
| 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%† | – |
| SkillsBench | Vals AI | 46.5%† | – |
| SREBench | Vals AI | 0.4%† | – |
| Tax Agent Bench | Vals AI | 29.4%† | – |
| Terminal-Bench 4.0 (Vals) | Vals AI | 10.6%† | – |
| Terminal-Bench Science | Vals AI | 5.7%† | – |
| Arena Search | Arena (LMArena) | 1191† | 1197 |
| Arena Agent Task Completion | Arena (LMArena) | -0.6 | – |
| DeepResearch Bench | DeepResearch Bench | – | 45.2%† |
| METR Time Horizons | METR | – | 63.6% |
| GBAEval | Epoch AI Benchmarking Hub | 65.3%† | – |
| Vending-Bench 2 | Andon Labs | 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 | 1222† | – |
| BrowseComp-Plus | BrowseComp-Plus | – | 50.5% |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 65.6%† | 33.3%† |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 29.3%† | – |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 80.0%† | 83.9%† |
| MATH Level 5 | Epoch AI Benchmarking Hub | – | 97.8%† |
| ProofBench | Vals AI | 77.0%† | – |
| LiveBench Mathematics | LiveBench | 92.9%† | – |
| AIME (Vals) | Vals AI | – | 85.3%† |
| MGSM | Vals AI | – | 91.8%† |
| IMO 2025 | MathArena | – | 16.7%† |
| Omni-MATH (HELM) | HELM Capabilities | – | 71.4% |
| MATH 500 | Vals AI | – | 94.6%† |
| Arena Maths | Arena (LMArena) | 1472 | 1448 |
| AIME 2025 (AA) | Artificial Analysis | – | 88.3% |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 33.7%† | 49.4%† |
| LiveBench 1.0 Data Analysis | LiveBench | 71.7%† | – |
| MMLU-Pro | Vals AI | 87.5%† | 85.6%† |
| LegalBench | Vals AI | 83.9%† | 83.8%† |
| CorpFin | Vals AI | 67.0%† | 59.7%† |
| TaxEval | Vals AI | 75.6%† | 74.6%† |
| MedQA | Vals AI | – | 96.1%† |
| PRBench Finance | Scale AI SEAL | – | 47.7% |
| PRBench Legal | Scale AI SEAL | – | 48.6% |
| MultiNRC | Scale AI SEAL | – | 45.5%† |
| MMLU-Pro (HELM) | HELM Capabilities | – | 85.9% |
| Excel Modeling Benchmark | Vals AI | 66.3%† | – |
| MedCode | Vals AI | 47.5%† | 47.3%† |
| MedScribe | Vals AI | 76.0%† | 76.7%† |
| Public Benefits Bench v1.1 | Vals AI | 66.0%† | – |
| DrugDiscoveryBench | Scale AI SEAL | 50.0%† | – |
| FACTS Parametric | 20.5%† | – | |
| Instruction following | |||
| LiveBench Language | LiveBench | 75.0%† | – |
| LiveBench Instruction Following | LiveBench | 63.9%† | – |
| IFBench | Artificial Analysis | – | 71.4% |
| MultiChallenge | Scale AI SEAL | – | 56.6%† |
| TutorBench | Scale AI SEAL | – | 52.1%† |
| IFEval (HELM) | HELM Capabilities | – | 86.9% |
| WildBench (HELM) | HELM Capabilities | – | 86.1% |
| Arena Instruction Following | Arena (LMArena) | 1467 | 1403 |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1461 | 1432 |
| EQ-Bench 4 | EQ-Bench | 1221† | – |
| EQ-Bench Creative Writing v3 | EQ-Bench | 1798† | 1675 |
| Arena Creative Writing | Arena (LMArena) | 1434 | 1383 |
| Arena Multi-turn | Arena (LMArena) | 1473 | 1420 |
| Arena Longer Queries | Arena (LMArena) | 1481 | 1411 |
| Lech Mazur Creative Writing | Lech Mazur | – | 83.9%† |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1272 | 1214 |
| MMMU-Pro | Artificial Analysis | 77.3%† | 70.1% |
| VISTA | Scale AI SEAL | – | 50.1%† |
| MMMU-Pro (Vals) | Vals AI | 83.0%† | 80.4%† |
| MortgageTax | Vals AI | 70.0%† | 65.7%† |
| Humanity's Sixth Sense | Scale AI SEAL | 24.8%† | – |
| VTB | Scale AI SEAL | – | 13.7% |
| GeoBench | GeoBench | 74.0% | 60.0%† |
| VPCT | Epoch AI Benchmarking Hub | – | 52.0%† |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 24.9%† | – |
| Long context | |||
| Fiction.LiveBench 120k | Fiction.live | – | 100.0%† |
| AA-LCR v1.1 | Artificial Analysis | 76.7% | 74.7% |
| Arena Document | Arena (LMArena) | 1469 | – |
| MLCR | Artificial Analysis | 4.4% | – |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 156.9† | 147.1 |
| LiveBench 1.0 | LiveBench | 76.0%† | – |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 31.7 | 20.2 |
| AA Progress Index v1 | Artificial Analysis | 296.8 | 258.3 |
| HELM Capabilities mean | HELM Capabilities | – | 81.1% |
| Vals Multimodal Index | Vals AI | 68.8%† | – |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | – | 16.0% |
| MASK | Scale AI SEAL | – | 84.5%† |
| PropensityBench | Scale AI SEAL | – | 10.5% |
| DistressBench | Scale AI SEAL | 68.0%† | – |
| Metadata | |||
| AA Openness Index | Artificial Analysis | – | 5.6 |
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
- o3: BenchLeader Index 61.2 → 60.4.
- o3: 1 score revision.
- 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.
- o3: 1 new result (Arena Search).
- o3: 1 score revision.
- Claude Sonnet 5 (high): 1 score revision.
- Claude Sonnet 5 (high): 1 new result (AA Progress Index v1).
Related comparisons
- Claude Sonnet 5 vs Gemini 4 Carbon
- Claude Sonnet 5 vs Claude Opus 5.5
- Claude Sonnet 5 vs GPT-6 Astra
- o3 vs Claude Opus 5.5
- Claude Sonnet 5 vs GPT-6.1 Sol
- Claude Sonnet 5 vs Gemini 4 Argon
- Claude Sonnet 5 vs Muse Spark
- Claude Sonnet 5 vs GPT-5.5
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
Claude Sonnet 5 vs o3: questions
- Is Claude Sonnet 5 better than o3?
- Claude Sonnet 5 (high) and o3 are neck and neck on quality (59.0 vs 60.4, 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 Sonnet 5 vs o3: which is cheaper, and will that change?
- o3 is cheaper today: $3.50 against $4.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is Claude Sonnet 5 or o3 better for coding?
- They are level in coding (60 vs 60 on the category index). On Arena Coding, the top shared coding benchmark, it is 1520 for Claude Sonnet 5 against 1460 for o3.
- Which is faster, Claude Sonnet 5 or o3?
- o3 streams faster (131 vs 99 output tokens per second, Artificial Analysis 7-day median).
- Is Claude Sonnet 5 or o3 better for agents and tool use?
- o3 scores higher in agentic tasks (70 vs 53 on the category index, where 50 is average).
More questions (4)
- Which is better for long documents?
- Claude Sonnet 5 scores higher on long-context benchmarks (63 vs 62). Claude Sonnet 5 accepts more context: 1,000,000 against 200,000 tokens for o3.
- Which has lower latency?
- o3 answers first: 5.2 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 Sonnet 5 vs o3?
- Across the 13 benchmarks both have, Claude Sonnet 5 leads 13, o3 leads 0. Every overall index that rates both puts Claude Sonnet 5 ahead.
- When was Claude Sonnet 5 released?
- Claude Sonnet 5 was released on 30 Jun 2026.