Claude Sonnet 5 vs Claude Sonnet 5.5
Different efforts: medium · high · Compare all at high
Humanity's Last Exam Humanity's Last Exam: Very hard expert questions across many fields, written to stump today’s models
Sonnet 530
Sonnet 5.545.8
Arena Text · overall Arena Text · overall: Human pairwise preferences in the stated text vote pool
Sonnet 5 · high1,461
Sonnet 5.5 · xhigh1,475
Arena Text · Chinese Arena Text · chinese: Human pairwise preferences in the stated text vote pool
Sonnet 5 · high1,477.6
Sonnet 5.5 · xhigh1,536.9
Arena Text · English Arena Text · english: Human pairwise preferences in the stated text vote pool
Sonnet 5 · high1,450.1
Sonnet 5.5 · xhigh1,484.5
Arena Text · French Arena Text · french: Human pairwise preferences in the stated text vote pool
Sonnet 5 · high1,463.8
Sonnet 5.5 · xhigh1,496.9
Arena Text · instruction following Arena Text · instruction following: Human pairwise preferences in the stated text vote pool
Sonnet 5 · high1,452.4
Sonnet 5.5 · xhigh1,488.5
Arena Text · non-English Arena Text · non english: Human pairwise preferences in the stated text vote pool
Sonnet 5 · high1,431.2
Sonnet 5.5 · xhigh1,454.2
Arena Text · Russian Arena Text · russian: Human pairwise preferences in the stated text vote pool
Sonnet 5 · high1,449.7
Sonnet 5.5 · xhigh1,445.2
Arena Text · Spanish Arena Text · spanish: Human pairwise preferences in the stated text vote pool
Sonnet 5 · high1,437.2
Sonnet 5.5 · xhigh1,450.2
AA Omniscience Index AA Omniscience Index: Scores factual knowledge while penalizing incorrect answers to encourage admitting uncertainty
Sonnet 5-6.9
Sonnet 5.521
SimpleQA Verified v1 SimpleQA Verified v1: Accuracy on short factual questions with verified answers, answered without web access
Sonnet 5 · max33.7
Sonnet 5.5 · max46.5
CritPt (physics) CritPt (physics): Research-level physics problems that need many careful reasoning steps
Sonnet 58.6
Sonnet 5.524.6
AA Intelligence Index 4.3.2 AA Intelligence Index 4.3.2: Artificial Analysis combines tests of knowledge, reasoning, coding and tool use into one score
Sonnet 528.1
Sonnet 5.546.8
AA Omniscience · Accuracy AA Omniscience · Accuracy: Measures correct answers on the AA Omniscience factual knowledge test
Sonnet 537.1
Sonnet 5.552
AA Omniscience · Hallucination Rate AA Omniscience · Hallucination Rate: Measures how often attempted answers are incorrect on the AA Omniscience factual knowledge test · lower is better
Sonnet 569.9
Sonnet 5.564.6
AA-LCR (long context) 1.1 AA-LCR (long context) 1.1: Questions that need reading and connecting facts across very long documents (around 100K tokens)
Sonnet 573.7
Sonnet 5.578
GDP.pdf · AA all-pass GDP.pdf · AA all-pass: Models reason over professional PDF documents; an attempt passes only when every atomic criterion passes
Sonnet 511.6
Sonnet 5.525.2
MLCR-AA MLCR-AA: Models synthesize long medical records; the headline credits concise answers judged both accurate and complete
Sonnet 5 · max55
Sonnet 5.5 · max75
Chess Puzzles Chess Puzzles: Models choose the best next move from a novel chess position
Sonnet 5 · max45
Sonnet 5.5 · max59
FrontierMath Tier 4 (v2) FrontierMath Tier 4 (v2): Models solve expert-written advanced mathematics problems on the named tier and edition
Sonnet 5 · max29.3
Sonnet 5.5 · max80.5
FrontierMath Tiers 1–3 (v2) FrontierMath Tiers 1–3 (v2): Models solve expert-written advanced mathematics problems on the named tier and edition
Sonnet 5 · max65.6
Sonnet 5.5 · max88.8
Mystery Game Puzzles Mystery Game Puzzles: Models choose the best move in puzzle positions from an undisclosed game
Sonnet 5 · max35
Sonnet 5.5 · max65
Vals MedCode Vals MedCode: Medical coding accuracy on patient encounters
Sonnet 547.5
Sonnet 5.552.9
Vals MedScribe Vals MedScribe: Accuracy of clinical notes produced from consultation transcripts
Sonnet 576.1
Sonnet 5.591.1
Vals Legal Research Bench Vals Legal Research Bench: Grounded legal research; a task passes only if all criteria pass
Sonnet 541.8
Sonnet 5.548.1
Arena Text · German Arena Text · german: Human pairwise preferences in the stated text vote pool
Sonnet 5 · high1,441.9
Sonnet 5.5—
Arena Text · Japanese Arena Text · japanese: Human pairwise preferences in the stated text vote pool
Sonnet 5 · high1,424.8
Sonnet 5.5—
Arena Text · Korean Arena Text · korean: Human pairwise preferences in the stated text vote pool
Sonnet 5 · high1,413
Sonnet 5.5—
Arena Text · Polish Arena Text · polish: Human pairwise preferences in the stated text vote pool
Sonnet 5 · high1,426.4
Sonnet 5.5—
FrontierMath Erdos FrontierMath Erdos: Models solve expert-written advanced mathematics problems on the named tier and edition
Sonnet 5—
Sonnet 5.5 · max2.9
Vals LegalBench Vals LegalBench: Legal reasoning and legal-language understanding on LegalBench
Sonnet 583.9
Sonnet 5.5—
Vals Science Index Vals Science Index: Scientific problem solving across the index’s component evaluations
Sonnet 5—
Sonnet 5.559.1
| Benchmark | Sonnet 5 | Sonnet 5.5 |
|---|---|---|
| Humanity's Last Exam | 30 | 45.8 |
| Arena Text · overall | 1,461 · high | 1,475 · xhigh |
| Arena Text · Chinese | 1,477.6 · high | 1,536.9 · xhigh |
| Arena Text · English | 1,450.1 · high | 1,484.5 · xhigh |
| Arena Text · French | 1,463.8 · high | 1,496.9 · xhigh |
| Arena Text · instruction following | 1,452.4 · high | 1,488.5 · xhigh |
| Arena Text · non-English | 1,431.2 · high | 1,454.2 · xhigh |
| Arena Text · Russian | 1,449.7 · high | 1,445.2 · xhigh |
| Arena Text · Spanish | 1,437.2 · high | 1,450.2 · xhigh |
| AA Omniscience Index | -6.9 | 21 |
| SimpleQA Verified v1 | 33.7 · max | 46.5 · max |
| CritPt (physics) | 8.6 | 24.6 |
| AA Intelligence Index 4.3.2 | 28.1 | 46.8 |
| AA Omniscience · Accuracy | 37.1 | 52 |
| AA Omniscience · Hallucination Rate | 69.9 | 64.6 |
| AA-LCR (long context) 1.1 | 73.7 | 78 |
| GDP.pdf · AA all-pass | 11.6 | 25.2 |
| MLCR-AA | 55 · max | 75 · max |
| Chess Puzzles | 45 · max | 59 · max |
| FrontierMath Tier 4 (v2) | 29.3 · max | 80.5 · max |
| FrontierMath Tiers 1–3 (v2) | 65.6 · max | 88.8 · max |
| Mystery Game Puzzles | 35 · max | 65 · max |
| Vals MedCode | 47.5 | 52.9 |
| Vals MedScribe | 76.1 | 91.1 |
| Vals Legal Research Bench | 41.8 | 48.1 |
| Arena Text · German | 1,441.9 · high | — |
| Arena Text · Japanese | 1,424.8 · high | — |
| Arena Text · Korean | 1,413 · high | — |
| Arena Text · Polish | 1,426.4 · high | — |
| FrontierMath Erdos | — | 2.9 · max |
| Vals LegalBench | 83.9 | — |
| Vals Science Index | — | 59.1 |
Specs
| Spec | Sonnet 5 | Sonnet 5.5 |
|---|---|---|
| Input, per 1M | $2 | $2 |
| Output, per 1M | $10 | $10 |
| Cost per task? | $1.00 | $0.88 |
| Speed, tok/s | 66 | 106 |
| Context | 1M | 1M |
| Weights | Closed | Closed |
| Released | Jun 2026 | Sep 2026 |
| See images | ✓ | ✓ |
| Read PDFs | Partial | ✓ |
| Hear audio | — | — |
| Watch video | — | — |
| Call tools | ✓ | ✓ |
| Force a tool call | Partial | — |
| Return strict JSON | ✓ | ✓ |
| Search the web | ? | ✓ |
| Run code | ? | ✓ |
| Use a computer | ? | ✓ |
What breaks when you switch
Sonnet 5 → Sonnet 5.5
No forced tool use›
Forced tool_choice any or tool returns HTTP 400; remove the forced choice.
Replace disabled thinking›
Disabled thinking returns HTTP 400; use between_tools for the lowest thinking setting.
No manual thinking budgets›
Manual thinking budgets and extra between_tools fields return HTTP 400; remove manual budgets and send between_tools without display, budget_tokens or block_binding.
Remove sampling parameters›
Non-default temperature, top_p or top_k returns HTTP 400; remove these parameters.
No assistant prefill›
A prefilled final assistant turn returns HTTP 400; remove that prefill.
Old computer tool rejected · Claude API / Google Cloud›
The computer_20251124 tool returns HTTP 400 on the Claude API and Google Cloud; use computer_toolset_20260801 there.
Sonnet 5.5 → Sonnet 5
Nothing known.