Inference
EN

Claude Opus 5.5 vs Step 5 Preview

Change models
Claude Opus 5.5Anthropic
Step 5 PreviewStepFun
IndexAll categories, out of 100.
78.6 ±2.4 best
68.4 ±3.7
GapIndex points below the top.
Top
−10.2
CoverageIndex weight that has results.
95% best
90%
SpeedOutput tokens per second.
76/s
105/s best
EloLMArena rating and rank.
1513 best
1443
ReleasedMonth of release.
2026-09
2026-09
Input / 1MUS dollars per 1M input tokens.
$4.00
$1.00 best
Output / 1MUS dollars per 1M output tokens.
$20.00
$2.70 best
Batch output / 1MBatch price per 1M output tokens.
$10.00
—
1K chats2K tokens in, 500 out, each.
$18.00
$3.35 best
1K documents20K tokens in, 1K out, each.
$100.00
$22.70 best
1K agent tasks200K tokens in, 20K out, each.
$1,200.00
$254.00 best
ContextMaximum tokens in one request.
1M
1M
AgenticMulti-step tasks with tools.
74.1 best
68.8
CodingCode writing and repair.
81.8 best
71.4
ReasoningLogic problems and puzzles.
80.8 best
67.4
MultimodalTasks with images and text.
74.1 best
62.2
KnowledgeFacts and expert knowledge.
79.2 best
63.5
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
59.6
—
MathMath problems.
77.7 best
61.7

Benchmarks

47 shared
Only Claude Opus 5.5: 110Only Step 5 Preview: 15
Claude Opus 5.5
Step 5 Preview
Artificial Analysis Humanity's Last ExamKnowledge
61.4% best2 / 222
46.5%23 / 222
Artificial Analysis Long Context ReasoningReasoning
84.7%6 / 217
88.3% best2 / 217
Critical Physics TasksReasoning
31.7% best4 / 214
20.9%26 / 214
Artificial Analysis Omniscience AccuracyKnowledge
65.5% best3 / 164
36.4%64 / 164
GDPval-AA normalizedAgentic
68.3% best1 / 133
54.3%24 / 133
Vibe Code Bench v1.1Coding
90.3% best6 / 123
70.4%48 / 123
Artificial Analysis SciCodeCoding
66.9% best2 / 114
58.9%13 / 114
Artificial Analysis MMMU-ProMultimodal
87.7% best1 / 113
76.4%44 / 113
Legal Research BenchLegal
50.5% best7 / 90
40.4%35 / 90
Tax Agent BenchFinance
44.5% best6 / 82
27.3%39 / 82
LiveBench Data AnalysisReasoning
80.3% best9 / 80
63.5%69 / 80
LiveBench ReasoningReasoning
92.2% best4 / 80
83.8%44 / 80
Finance Agent (v2)Finance
58.6% best10 / 79
50.7%42 / 79
Code MigrationCoding
66.6% best4 / 77
39.0%31 / 77
EMBFinance
75.9% best2 / 75
57.1%41 / 75
Terminal-Bench 4.0Agentic
65.2% best1 / 61
31.8%23 / 61
DeepSWEAgentic
74.2% best10 / 59
67.7%27 / 59
Agent Arena steerabilityAgentic
18.6 best3 / 54
−0.930 / 54
Agent Arena task outcomeAgentic
20.7 best2 / 54
−1.535 / 54
ProofBench v1.1Math
100.0% best1 / 54
42.0%37 / 54