Skip to content

Claude Opus 5.5

AnthropicProprietaryReasoningVerifiedReleased Sep 22, 2026Input: text, image, file
Compare with GPT-6.1 SolSource scores ↗OpenRouter ↗
Overall score
86.7
Rank 1 of 222 · 81.8–91.6
Context window
1M
1,000,000 tokens
Price per 1M tokens
$4.00 in · $20.00 out
✓ Verified on Anthropic's pricing page
Blended price
$8.00
3:1 input:output tokens
Output speed
—
Not measured yet

Category scores

Scores per capability area with the model's rank among ranked models in that category.

  1. 1Math97.3
  2. 2Multimodal#392.2
  3. 3Coding#287.5
  4. 4Reasoning#486.1
  5. 5Knowledge#185.4
  6. 6Agentic & tool use#165.3
Higher is better. #N is the model's rank in that category, where it is eligible to be ranked.

Benchmark results (64)

As reported in system cards, papers and public evaluations. Position compares this model with every other model reporting the same benchmark.

Knowledge

HLE w/o tools64.4#1 of 43
HealthBench Professional65.6#2 of 16
HealthBench (raw)68.1#2 of 7
HealthBench (length-adjusted)60.6#2 of 7
HealthBench Professional (raw)77.1#1 of 7
BioMysteryBench (human-difficult)50.0#2 of 6
BioMysteryBench (human-solvable)89.3#2 of 5
SpatialBench Verified72.0#3 of 3
SingleCellBench61.2#1 of 3
Protein Design60.2#1 of 3
Protocols (troubleshooting)73.7#1 of 3
Protocols (understanding)69.0#2 of 3
Morphology-to-molecule matching34.0#1 of 2
Medicinal chemistry63.5#2 of 2
Protein Design library ranking56.0#1 of 2
De novo protein-binder design82.6#1 of 2

Agentic & tool use

MazeBench13#3 of 42
HLE w/ tools67.7#1 of 27
AutomationBench40.0#14 of 25
OSWorld 2.048.7#10 of 24
Toolathlon-Verified77.8#3 of 22
Terminal-Bench 4.066.40#2 of 18
Hermes Index63.3#1 of 16
Hermes Bench76.7#1 of 16
CWE-bench v167.0#4 of 11
Terminal-Bench-Science 0.158.7#3 of 5
Toolathlon Verified Pass@382.4#3 of 4
Toolathlon Verified Pass³72.2#3 of 4
Toolathlon Verified avg. turns(lower is better)26.9#3 of 4
LAB all-pass (Harvey held-out)8.3#3 of 3
LAB criterion-pass (Harvey held-out)91.2#3 of 3

Multimodal

Design Arena Website1335#4 of 91
ParseBench79.85#1 of 37
ExtractBench (LlamaIndex)(lower is better)86.06#10 of 24
OfficeQA Pro67.7#1 of 13
SWE-bench Multimodal61.4#1 of 7
ExtractBench Pro51#2 of 5
Chartography (tools)89.0#2 of 5
Chartography (no tools)64.4#2 of 5
BenchCAD Vision2Code (tools)0.962#2 of 4
BenchCAD Vision2Code (no tools)0.730#2 of 3
OfficeQA78.9#1 of 3
Biomedical image analysis71.4#2 of 2

Coding

SWE-bench Pro89.9#1 of 69
DeepSWE74.2#7 of 50
Bug Hunt Bench41.7#7 of 23
FrontierSWE v262.3#2 of 20
FrontierCode 1.1 Main54.4#1 of 16
ProgramBench91.2#2 of 15
PostTrainBench v1.149.3#1 of 15
FrontierCode 1.1 Extended63.6#2 of 8

Reasoning

ARC-AGI-291.7#4 of 41
ARC-AGI-197.50#4 of 36
STEMBench2.62#4 of 9
GraphWalks BFS 256K–1M66.8#3 of 4

Math

FrontierMath v2 (Tiers 1-3)91.228#3 of 64
FrontierMath v2 (Tier 4)95.000#3 of 51
ArXivMath Aug. 2026 (no tools)91.2#1 of 2
ArXivMath Aug. 2026 (tools)96.9#1 of 2

Multilingual

SWE Multilingual93.9#1 of 46
GMMLU94.3#1 of 3
MILU93.1#1 of 3

Instruction following

Gray Swan IPI (15 attempts)(lower is better)1.0#2 of 12

Pricing by source

USD per 1M tokens.

SourceInputOutputCached in
✓ Official (Anthropic)
checked Oct 17, 2026 · Claude Opus 5.5
$4.00$20.00$0.20
BenchLM list price$4.00$20.00$0.20
OpenRouter anthropic/claude-opus-5.5$4.00$20.00$0.20

Anthropic's September 22, 2026 Claude Opus 5.5 model documentation lists the `claude-opus-5-5` API model at $4 input / $20 output per million tokens, with $0.20 cache reads. Five-minute and one-hour cache writes are $5 and $8 per million tokens; Batch API input and output receive a 50% discount.

Arena ratings

Human-preference ratings from LMArena. Best-rated variant shown.

Text(High)1516#2 · 8.3K votes
WebDev (coding)(Max)1807#1 · 2.9K votes
Vision(High)1321#2 · 3K votes
Search(medium)1213#14 · 10.8K votes

Speed

Measured by us through OpenRouter (median of trials).

Not measured yet.

Related models

Same family

Similar score