Claude Sonnet 5 vs o3
Claude Sonnet 5 
o3 Verdict
- Claude Sonnet 5 (high) and o3 are level on quality (61.3 vs 61.6).
- Claude Sonnet 5 (high) is stronger in composite, human preference, knowledge, long context, multimodal, reasoning.
- o3 is stronger in agents & tools, coding, instruction following, maths.
- They cost about the same ($3.50 per 1M blended).
- o3 streams 1.8× faster (107 vs 61 tokens per second).
| Metric | Claude Sonnet 5 (high) | o3 |
|---|---|---|
| BenchLeader Index | 61.3 | 61.6 |
| Agents & tools score | 60.8 | 70.3 |
| Coding score | 61.5 | 62.0 |
| Composite score | 69.5 | 54.5 |
| Human preference score | 65.9 | 62.3 |
| Knowledge score | 62.5 | 58.0 |
| Long context score | 64.8 | 63.8 |
| Multimodal score | 63.1 | 57.9 |
| Reasoning score | 66.5 | 61.5 |
| Instruction following score | – | 65.0 |
| Maths score | – | 78.6 |
| Blended price $/M | $4.00 | $3.50 |
| Output speed | 61 tok/s | 107 tok/s |
| Time to first answer | 2.0 s | 6.3 s |
| Context window | 1M | 200k |
| SciCode | 48.6% | – |
| WeirdML | 68.8% | – |
| Epoch Capabilities Index | – | 146.9 |
| LMArena Text | 1461 | 1432 |
| LMArena Hard Prompts | 1488 | 1441 |
| LMArena Coding | 1521 | 1460 |
| LMArena WebDev | 1537 | – |
| LMArena Vision | 1278 | 1214 |
| LMArena Agent | 5.9 | – |
| AA Intelligence Index | 32.0 | 20.2 |
| IFBench | – | 71.4% |
| AA-LCR | 76.7% | 74.7% |
| MMMU-Pro | – | 70.1% |
| AA-Omniscience | -3.7 | -15.6 |
| Terminal-Bench Hard | – | 37.1% |
| GPQA Diamond (AA) | – | 82.7% |
| Humanity's Last Exam (AA) | 35.7% | 20.1% |
| SciCode (AA) | 54.3% | – |
| τ²-Bench Telecom (AA) | – | 80.7% |
| PRBench Finance | – | 47.7% |
| PRBench Legal | – | 48.6% |
| MMMU (validation) | – | 82.9% |
| MMMU-Pro (official) | – | 76.4% |
| Kagi LLM Benchmark | – | 67.6% |
| IFEval (HELM) | – | 86.9% |
| Omni-MATH (HELM) | – | 71.4% |
| WildBench (HELM) | – | 86.1% |
| MMLU-Pro (HELM) | – | 85.9% |
| GPQA Diamond (HELM) | – | 75.3% |
| HELM Capabilities mean | – | 81.1% |
| Aider Polyglot | – | 81.3% |
| SWE-bench Verified (bash only) | – | 58.4% |
| SWE-bench Verified (any scaffold) | – | 58.4% |
| BFCL Overall | – | 63.0% |
Data as of 2026-09-12. Best configuration of each model; every score links to its source on the model pages.
Claude Sonnet 5 vs o3: questions
- Is Claude Sonnet 5 better than o3?
- Claude Sonnet 5 (high) and o3 are level on quality (61.3 vs 61.6). The BenchLeader Index combines every independent quality benchmark; o3 is ahead overall as of 2026-09-12, but check the category scores for your use.
- Is Claude Sonnet 5 better than o3 for coding?
- o3 scores higher in coding (62 vs 62 on the category index, where 50 is average).
- Is Claude Sonnet 5 better than o3 for agentic tasks?
- o3 scores higher in agentic tasks (70 vs 61 on the category index, where 50 is average).
- Which is cheaper, Claude Sonnet 5 or o3?
- o3 is cheaper: $3.50 against $4.00 per million tokens, blended at three input tokens per output token.
- Which is faster, Claude Sonnet 5 or o3?
- o3 streams faster: 107 against 61 output tokens per second.
- Which has the larger context window?
- Claude Sonnet 5 accepts more context: 1M against 200k tokens.