Claude Sonnet 5.5

Claude Sonnet 5.5 ranks first on 10 of the 287 benchmarks with five or more models scored.

Anthropic Released September 28th, 2026 Family Claude 5.5 The lab's model card Compare with others

298

Benchmarks with a score

Ranked first, out of five or more models

In the top three

Anthropic Sep 2026

Claude Sonnet 5.5

Closed weights Mid-size

166.6 #8 of 256

Epoch Capabilities Index

Self-Reported Index 168.0 #7 of 404

Humanity's Last Exam 46.1% #4 /47
AutomationBench 44.8% #5 /42
Terminal-Bench 4.0 61.8% #2 /40

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

Benchmark	Score	Rank	Source
Humanity's Last Exam	46.1%	#4 of 47	Official leaderboard
LMArena (Elo)	1,472	#39 of 335	Official leaderboard
HANDBOOK.md	47.7%	#2 of 16	Official leaderboard
Joke Arena	1,550	#4 of 25	Official leaderboard
AA-LCR Extended	73.8%	#8 of 49	Official leaderboard
ComplexConstraints	67.2%	#4 of 15	Official leaderboard
BullshitBench V2	60%	#35 of 129	Official leaderboard
Corp Speak Arena	1,562	#7 of 22	Official leaderboard
SimpleQA Verified	44.8%	#19 of 56	Official leaderboard
Global MMLU	92.1%	#5 of 14	Lab model card
Global MMLU Lite	92.6%	#16 of 43	Official leaderboard
Extended NYT Connections	80.5%	#41 of 103	Official leaderboard
HLE-Diamond	43.3%	#8 of 20	Official leaderboard
FACTS Multimodal	50.4%	#10 of 23	Official leaderboard
MILU	91.6%	#5 of 11	Lab model card
FACTS Grounding v2	62.5%	#16 of 25	Official leaderboard
FACTS Benchmark Suite	58.7%	#18 of 23	Official leaderboard
Opening Lines Arena	1,425	#12 of 15	Official leaderboard
FACTS Parametric V2	35.1%	#22 of 27	Official leaderboard
ELI5 Arena	1,407	#16 of 18	Official leaderboard
StudyArena Text	1,170	#79 of 85	Official leaderboard

Math and science

Benchmark	Score	Rank	Source
GPQA Diamond	95.6%	#4 of 229	Epoch AI run
OTIS Mock AIME 2024-2025	100%	#10 of 212	Epoch AI run
DTBench	97.3%	#10 of 180	Official leaderboard
ProofBench v1.1	100%	#3 of 52	Official leaderboard
BioMysteryBench	81.1%	#2 of 30	Official leaderboard
LatchBio Capabilities v1.0	54.2%	#2 of 27	Official leaderboard
FrontierMath Tiers 1-3 v2	88.8%	#10 of 86	Official leaderboard
MysteryMechanism	49.1%	#4 of 31	Official leaderboard
ARC-AGI	96.5%	#15 of 95	Official leaderboard
FrontierMath Tier 4 v2	80.5%	#11 of 68	Official leaderboard
SimpleBench	75.9%	#18 of 106	Official leaderboard
ARC-AGI-2	83.9%	#18 of 98	Official leaderboard
scBench (LatchBio SingleCellBench)	60.3%	#6 of 26	Official leaderboard
Riemann-bench	76.8%	#6 of 20	Official leaderboard
FrontierMath Erdős	2.90%	#5 of 15	Official leaderboard
Terminal-Bench-Science	59.9%	#4 of 12	Lab model card
Biomedical image analysis (Anthropic)	72.2%	#1 of 3	Lab model card
Medicinal Chemistry (Anthropic)	65.3%	#1 of 3	Lab model card
Prophet Arena	85.8	#9 of 24	Official leaderboard
SpatialBench (LatchBio)	72.5%	#3 of 7	Lab model card
ArXivMath	86.8%	#4 of 9	Lab model card
De novo protein binder design (Anthropic)	82.3%	#3 of 3	Lab model card
Morphology-to-molecule matching (Anthropic)	25%	#3 of 3	Lab model card

Coding

Benchmark	Score	Rank	Source
Code Migration	69.8%	#1 of 78	Official leaderboard
Vibe Code Bench	92.4%	#2 of 120	Official leaderboard
DesignArena Data Visualization	1,360	#2 of 118	Official leaderboard
DesignArena UI Component	1,386	#3 of 139	Official leaderboard
Code Arena	1,773	#3 of 128	Official leaderboard
MirrorCode	97.8%	#1 of 39	Epoch AI run
SWE Atlas Codebase QnA Extended	44.7%	#1 of 38	Official leaderboard
WebDev Arena	1,786	#3 of 105	arena.ai, via Epoch AI
APEX-SWE	66.4%	#2 of 65	Official leaderboard
Code Migration v1.1	86.9%	#1 of 30	Official leaderboard
Image-to-WebDev Arena	1,744	#2 of 55	Official leaderboard
SWE-Bench Pro	81.3%	#3 of 81	Lab model card
SWE-bench Multilingual	90.3%	#3 of 73	Lab model card
DesignArena Game Dev	1,391	#8 of 143	Official leaderboard
Android Bench 2.0	40%	#1 of 17	Official leaderboard
DesignArena 3D Design	1,401	#9 of 136	Official leaderboard
DesignArena Code	1,347	#11 of 143	Official leaderboard
DesignArena SVG	1,303	#9 of 103	Official leaderboard
DesignArena React Native	1,261	#5 of 56	Official leaderboard
ALE-Bench	2,448	#12 of 126	Official leaderboard
DesignArena Website	1,311	#15 of 151	Official leaderboard
DesignArena Android (Kotlin)	1,260	#5 of 46	Official leaderboard
sudo L7	50.6%	#2 of 18	Official leaderboard
DesignArena Agentic Overall	1,280	#9 of 74	Official leaderboard
DAYJOB: Data Science	35.3%	#2 of 15	Official leaderboard
FrontierCode 2.0	42.2%	#5 of 32	Official leaderboard
Brokk Power Ranking	26.9	#3 of 19	Official leaderboard
FrontierSWE	61.9%	#4 of 25	Official leaderboard
FDE-Gym	0.49	#2 of 12	Official leaderboard
DesignArena Frontend Web App	1,277	#9 of 53	Official leaderboard
CursorBench	55.5%	#4 of 21	cursor.com, via Epoch AI
CursorBench 4.0	55.5%	#4 of 21	Official leaderboard
IOI v2 (Vals)	83.1%	#9 of 45	Official leaderboard
SWE-Marathon v1.1	51.3%	#4 of 20	Official leaderboard
Real-SWE 1.1	51.2%	#3 of 15	Official leaderboard
FrontierCS 2	50%	#4 of 19	Official leaderboard
DesignArena Image to HTML	1,232	#11 of 51	Official leaderboard
SWE Atlas v1.1	68.3%	#7 of 28	Official leaderboard
DesignArena Image to Frontend	1,249	#8 of 31	Official leaderboard
OpenDesign Arena	86.8	#6 of 21	Official leaderboard
WeirdML v3	19.9%	#8 of 27	Official leaderboard
DAXBench v2	84.6%	#14 of 47	Official leaderboard
DeepSWE	71%	#19 of 56	Lab model card
OWE-Bench	39%	#5 of 14	Official leaderboard
FrontierCode 1.1	52.1%	#5 of 13	Lab model card
DesignArena Blender	1,268	#5 of 12	Official leaderboard
ViBench 2.0	19.1%	#4 of 9	Official leaderboard
SWE-Together	62.4%	#10 of 22	Official leaderboard
SWE-bench Verified Extended	57.3%	#21 of 44	Official leaderboard
SWE-bench Multimodal	54.3%	#6 of 12	Lab model card
RSI Bench	35.2	#4 of 8	Official leaderboard
FrontierCS 1.1	53.9	#3 of 6	Official leaderboard
SWE-Bench Pro V2	95.8%	#10 of 18	Official leaderboard
ProgramBench tests passed	79.7%	#7 of 10	Lab model card
SWE-Bench Pro V2 Hard	70.6%	#14 of 18	Official leaderboard
Chart Arena	1,457	#12 of 15	Official leaderboard
Floorplan Arena	1,257	#11 of 11	Official leaderboard
ReviewBench	63.1%	#4 of 4	Official leaderboard

Agents and tools

Benchmark	Score	Rank	Source
BrowseComp Extended	73%	#1 of 42	Official leaderboard
Terminal-Bench 4.0	61.8%	#2 of 40	Official leaderboard
LisanBench	13787.0	#9 of 133	Official leaderboard
Terminal-Bench 4.1	63.9%	#3 of 43	Official leaderboard
Agent Arena	15.9%	#4 of 56	Official leaderboard
Mystery Game Puzzles	65%	#7 of 84	Epoch AI run
o11y-bench	90.6%	#2 of 22	Official leaderboard
DuelLab GameBench 2	91.4	#6 of 57	Official leaderboard
AutomationBench	44.8%	#5 of 42	Official leaderboard
Chess Puzzles	59%	#12 of 93	Epoch AI run
LLM Chess	1,195	#16 of 123	Official leaderboard
Terminal-Bench 2.1 Extended	45.5%	#6 of 44	Official leaderboard
Cresta-Bench	78%	#2 of 14	Official leaderboard
Game Arena Bargaining	243	#6 of 39	Official leaderboard
Game Arena Nine Men's Morris	912	#7 of 39	Official leaderboard
Game Arena Dots and Boxes	1,055	#8 of 40	Official leaderboard
ARC-AGI-3	72.4%	#6 of 27	Official leaderboard
Game Arena Go	1,442	#9 of 39	Official leaderboard
Toolathlon-Verified	77.8%	#5 of 21	Lab model card
Collaborative Intelligence Benchmark	67.9%	#5 of 18	Official leaderboard
OSWorld 2.1	80.1%	#2 of 7	Lab model card
RuneBench	5.52	#24 of 76	Official leaderboard
Hermes Bench	73.5%	#6 of 17	Official leaderboard
Game Arena Checkers	980	#17 of 41	Official leaderboard
Game Arena Dark Hex	281	#17 of 41	Official leaderboard
DRACO	87%	#5 of 12	Lab model card
Game Arena Four in a Row	412	#20 of 43	Official leaderboard
Game Arena Reversi	540	#20 of 39	Official leaderboard
Game Arena Word Art	604	#20 of 39	Official leaderboard
Game Arena Chess	783	#22 of 42	Official leaderboard
Kaggle Game Arena	322	#22 of 42	Official leaderboard
BU Bench V2.1	72.7	#9 of 16	Official leaderboard
CivBench	1,497	#11 of 19	Official leaderboard
Game Arena Lines of Action	338	#24 of 39	Official leaderboard
Game Arena Ultimate Tic-Tac-Toe	245	#24 of 39	Official leaderboard
Game Arena Clobber	170	#26 of 39	Official leaderboard
StarSkirmish Bench	25.8%	#6 of 9	Official leaderboard
Game Arena Word Association	126	#29 of 40	Official leaderboard
WANDR	70%	#3 of 4	Lab model card
SpireClimb	1.33%	#8 of 9	Epoch AI run
SpireClimb StS2 A10	2.67%	#8 of 9	Official leaderboard
Game Arena Coin Game	200	#35 of 39	Official leaderboard
Web Context (Nimble)	67.2%	#10 of 11	Official leaderboard
SpireClimb StS1 A20 Heart	0%	#9 of 9	Official leaderboard

Professional work

Benchmark	Score	Rank	Source
Vals Index v2	69.2%	#2 of 65	Official leaderboard
MedScribe	91.1%	#3 of 97	Official leaderboard
HealthBench Professional	69.2%	#1 of 26	Lab model card
Excel Modeling Benchmark	75.7%	#3 of 76	Official leaderboard
Vals Index v2.1	67%	#2 of 49	Official leaderboard
Tuesday Work Index	75.8	#2 of 44	Official leaderboard
GDPval Extended	7.00%	#2 of 43	Official leaderboard
DAYJOB: Finance	40.5%	#2 of 39	Official leaderboard
DAYJOB: Healthcare	34%	#2 of 37	Official leaderboard
Vals Index v2.2	70.8%	#2 of 37	Official leaderboard
DAYJOB: Accounting	35.2%	#1 of 17	Official leaderboard
HealthBench	65.4%	#2 of 31	Lab model card
APEX-Agents	75.5%	#4 of 61	Collected by Epoch AI, source not named
DAYJOB: Tax	10.8%	#1 of 15	Official leaderboard
EnterpriseBench: CoreCraft	70.8%	#4 of 51	Official leaderboard
APEX-Agents 1.1	83.7%	#6 of 65	Official leaderboard
Excel Modeling Benchmark v1.1	75.9%	#4 of 42	Official leaderboard
Vending-Bench 2	$11,165	#8 of 67	Official leaderboard
PolicyBench	91.7%	#6 of 50	Official leaderboard
MedCode	52.9%	#12 of 96	Official leaderboard
Vals Finance Agent v2	58.1%	#11 of 80	Official leaderboard
Tax Agent Bench	42%	#12 of 84	Official leaderboard
DesignArena Spreadsheet	1,295	#5 of 35	Official leaderboard
Vals Finance Agent v2.1	67.3%	#6 of 40	Official leaderboard
Vals Finance Index	64.5%	#6 of 40	Official leaderboard
GDP.xlsx	29.1%	#3 of 20	Official leaderboard
RealmPathology	77.3%	#2 of 13	Official leaderboard
Legal Research Bench	48.1%	#15 of 93	Official leaderboard
EEBench	67.2%	#5 of 30	Official leaderboard
Physical-Engineering's Last Exam	0.45	#3 of 18	Official leaderboard
DeepFinance	54.6%	#4 of 23	Official leaderboard
MissionReport	63.8%	#2 of 10	Official leaderboard
APEX-Accounting	64.5%	#10 of 49	Official leaderboard
FailureSensorIQ	67.4%	#5 of 24	Official leaderboard
YC-Bench	$1,741,315	#13 of 59	Official leaderboard
RCM-Bench-1	91%	#5 of 20	Official leaderboard
AA-Briefcase v1.1	1,811	#2 of 8	Lab model card
DesignArena PDF	1,238	#10 of 37	Official leaderboard
Immigration Bench	69%	#7 of 25	Official leaderboard
Public Benefits Bench	67.2%	#15 of 50	Official leaderboard
GDPval-AA v2.1	1,844	#3 of 10	Lab model card
DAYJOB: Law	27%	#5 of 16	Official leaderboard
Clipping Bench	$886	#6 of 19	Official leaderboard
BenchCAD	74.7%	#4 of 11	Lab model card
micro1 Medical Index	70%	#3 of 8	Official leaderboard
Thrive Holdings helpdesk ticket resolution benchmark	51.9%	#6 of 15	Official leaderboard
TaxCalcBench v2	50%	#14 of 32	Official leaderboard
GDPx	8.22%	#9 of 20	Official leaderboard
OfficeQA	76.9%	#7 of 14	Lab model card
Legora BAR	102.0	#9 of 17	Official leaderboard
MedXpertQA MM Extended	55.6%	#22 of 41	Official leaderboard
Harvey LAB (Legal Agent Benchmark)	2.92%	#45 of 80	Official leaderboard
BenchCAD 2.0	41%	#3 of 5	Official leaderboard
BenchCAD 2.1	60.9%	#3 of 5	Official leaderboard
PharmaBench (micro1)	11.4%	#8 of 13	Official leaderboard
OfficeQA Pro	65.6%	#10 of 16	Lab model card
Harvey LAB v1.1 (HLAB-Vals-1.1)	10.4%	#24 of 38	Official leaderboard
Quant-Arena	40.9	#6 of 9	Official leaderboard
PhysicianBench	63.2%	#2 of 3	Lab model card
Vals Legal Index	68%	#27 of 38	Official leaderboard
HW-Bench	20%	#11 of 15	Official leaderboard
Short Form Bench	$2,355	#7 of 9	Official leaderboard
APEX-1	53.9%	#64 of 71	Official leaderboard
Company Agent Bench	68.7%	#6 of 6	Official leaderboard

Vision, audio and robotics

Benchmark	Score	Rank	Source
DesignArena Code to Video	1,318	#2 of 29	Official leaderboard
ExtractBench	94.7%	#2 of 27	Official leaderboard
MineBench	1,977	#8 of 75	Official leaderboard
Code2Video-Compass	1.93	#2 of 15	Official leaderboard
BVB	72.3	#8 of 45	Official leaderboard
CharXiv Extended	84.2%	#7 of 38	Official leaderboard
vCX Benchmark	91.2%	#8 of 42	Official leaderboard
Furniture Assembly Benchmark	75%	#7 of 35	Official leaderboard
MMMU-Pro Extended	44.4%	#8 of 38	Official leaderboard
GDP.pdf	51.2%	#4 of 19	Official leaderboard
Blueprint-Bench 2	39.3%	#9 of 38	Official leaderboard
Vision Arena	1,263	#36 of 136	Official leaderboard
ParseBench	69.2%	#13 of 44	Official leaderboard
ShadowBench	48.8	#2 of 6	Official leaderboard
Code2Video Arena	1,542	#5 of 13	Official leaderboard
ExtractBench Pro	32%	#3 of 7	Official leaderboard
VibeVideoBench	3.88	#4 of 9	Official leaderboard
SplatBench	5.70	#5 of 8	Official leaderboard
Dual-Arm Sim	26.0	#7 of 11	Official leaderboard
OmniParseBench	88.7	#6 of 9	Official leaderboard
Chartography	61.6%	#10 of 14	Lab model card

Safety and security

Benchmark	Score	Rank	Source
World-Class Bio	66.5%	#3 of 74	Official leaderboard
Human Pathogen Capabilities Test	77.2%	#4 of 86	Official leaderboard
BioSecBench-Function	53.5%	#2 of 27	Official leaderboard
scBench-Long	31.8%	#2 of 27	Official leaderboard
VariantBench	51.7%	#2 of 27	Official leaderboard
Prompt injection robustness against adaptive attackers, browser use (Anthropic)	0%	#1 of 13	Lab model card
BioTIER Refuse	95.7%	#6 of 69	Official leaderboard
OncologyBench	42.7%	#3 of 30	Official leaderboard
BioSecBench-Surveillance	54.6%	#3 of 27	Official leaderboard
EpiBench	30.2%	#3 of 27	Official leaderboard
MetagenomicsBench	60%	#3 of 26	Official leaderboard
Molecular Biology Capabilities Test	58.8%	#11 of 90	Official leaderboard
TxBench-Preclinical-Pharmacology	64.3%	#4 of 30	Official leaderboard
DeepsecBench	37.5	#4 of 29	Official leaderboard
VCT-v2	52.2%	#4 of 25	Official leaderboard
BioSimBench	0.16	#2 of 12	Official leaderboard
nocot-bench (NCRI)	95.1	#31 of 185	Official leaderboard
SAFE-Teen Urgent safety	23%	#3 of 17	Official leaderboard
SHADE-Arena	22.3%	#3 of 16	Lab model card
SRE-Bench	30.1%	#7 of 35	Official leaderboard
TxBench-Antibody-Discovery	57.3%	#6 of 30	Official leaderboard
SpatialBench-Long	36.5%	#6 of 27	Official leaderboard
TxBench-Oligo-Discovery	51.9%	#7 of 30	Official leaderboard
DistressBench	74.6%	#12 of 42	Official leaderboard
Prompt injection robustness against adaptive attackers, computer use (Anthropic)	0.07%	#4 of 14	Lab model card
VIALS 2	41%	#4 of 14	Official leaderboard
Virology Capabilities Test (VCT)	58%	#9 of 31	Lab model card
SAFE-Teen Relationship boundaries	37%	#5 of 17	Official leaderboard
Single-turn benign request evaluations (Anthropic)	0.02%	#6 of 19	Lab model card
Binary Exploitation Benchmark (Anthropic, OSS-Fuzz based)	85.8%	#2 of 6	Lab model card
Silent copying of answers (Anthropic)	8.00%	#1 of 3	Lab model card
CyScenarioBench	46.1%	#6 of 17	Lab model card
ExploitBench	80%	#8 of 21	Lab model card
ExploitGym	26.6%	#6 of 15	Lab model card
SAFE-Teen Health advice	43%	#7 of 17	Official leaderboard
WellnessBench	57%	#5 of 12	Official leaderboard
Anthropic political even-handedness eval	97.9%	#6 of 14	Lab model card
MASK	94.5%	#9 of 19	Lab model card
Self-preference bias (Anthropic)	0.16	#2 of 4	Lab model card
SecIT Bench	75.1%	#14 of 26	Official leaderboard
Gray Swan IPI Arena benchmark	0.40%	#6 of 11	Lab model card
Malicious use of Claude Code (Anthropic)	85.2%	#10 of 17	Lab model card
Suicide Memory Bench	0.65	#8 of 12	Official leaderboard
Anthropic undermining liberal democracy (behavioral audit metric)	1.21	#4 of 6	Lab model card
Concealment of changes (Anthropic)	96.2%	#2 of 3	Lab model card
Sandbox escape evaluation (Anthropic)	4.60%	#2 of 3	Lab model card
Prompt injection robustness against adaptive attackers, coding (Anthropic)	3.01%	#10 of 14	Lab model card
CyberBench v1.1 (Vals)	59.6%	#33 of 44	Official leaderboard
BBQ (Bias Benchmark for QA)	80%	#19 of 25	Lab model card
Biopharma Bench	49.3%	#10 of 13	Official leaderboard
Anthropic automated behavioral audit	2.87	#5 of 6	Lab model card
LinuxArena	36.2%	#5 of 6	Lab model card
Single-turn harmful request evaluations (Anthropic)	95.6%	#16 of 19	Lab model card
BioTIER Permit	78.5%	#65 of 69	Official leaderboard
Malicious computer use (Anthropic)	79.5%	#14 of 14	Lab model card
Rewind-attacker cyber adversarial robustness (Anthropic)	21%	#6 of 6	Lab model card
Cyber harm coverage (Anthropic)	99.4%	#5 of 5	Lab model card

Other

Benchmark	Score	Rank	Source
LMCA	68.3%	#2 of 147	Official leaderboard
ACCoRD	85.7	#3 of 152	Official leaderboard
Conceptual Reasoning Index	77.1	#3 of 151	Official leaderboard
SAGE (Vals)	51.8%	#11 of 81	Official leaderboard
Political Science LLM Benchmark	68.9%	#6 of 42	Official leaderboard
Vals Science Index	59.1%	#4 of 25	Official leaderboard
Hermes Index	53.1	#4 of 17	Official leaderboard
LiveBench	77%	#9 of 31	Official leaderboard
Anthropic ECI (AECI)	167.9	#2 of 6	Lab model card
SAGE v1.1 (Vals)	88%	#10 of 29	Official leaderboard

Its lab's numbers put it at 168.0, 1.4 points above its ECI

The Self-Reported Index fits one score to the 23 capability results Anthropic reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 165.8 to 170.2. How we compute it.

Its lab's numbers next to 7 leaderboards

Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.

Benchmark	Lab says	Board says	Gap
Humanity's Last Exam	64.5% Multidisciplinary reasoning Humanity's Last Exam (with tools)	46.1% Leaderboard, xhigh	+18 pts
Harvey LAB (Legal Agent Benchmark)	11.7% Legal Agent Benchmark held-out set, all-pass rate at high effort, mean criterion pass 92.1% (max effort 10% and 93.1%) (8.14.2)	2.92% Leaderboard	+8.8 pts
Terminal-Bench 4.0	70.6% Agentic coding Terminal-Bench 4.0	61.8% Leaderboard, max, Claude Code	+8.8 pts
scBench (LatchBio SingleCellBench)	59.1% LatchBio SingleCellBench (8.17.2, Figure 8.17.A)	60.3% Leaderboard, published v1, Pi	−1.2 pts
AutomationBench	44.7% AutomationBench	44.8% Leaderboard, max, default fallbacks	Same
FrontierSWE	61.9% FrontierSWE v2, mean of five trials, Proximal harness at max effort (8.7)	61.9% Leaderboard, mean@5, proximus	Same
CursorBench 4.0	55.5% Agentic coding CursorBench 4.0	55.5% Leaderboard, max, Cursor	Same