Inference
EN

Claude Fable 5.1 vs Step 5 Preview

Change models
Claude Fable 5.1Anthropic
Step 5 PreviewStepFun
IndexAll categories, out of 100.
76.2 ±2.4 best
68.4 ±3.7
GapIndex points below the top.
Top
−7.8
CoverageIndex weight that has results.
95% best
90%
SpeedOutput tokens per second.
69/s
105/s best
EloLMArena rating and rank.
1512 best
1443
ReleasedMonth of release.
2026-09
2026-09
Input / 1MUS dollars per 1M input tokens.
$10.00
$1.00 best
Output / 1MUS dollars per 1M output tokens.
$50.00
$2.70 best
Batch output / 1MBatch price per 1M output tokens.
$25.00
—
1K chats2K tokens in, 500 out, each.
$45.00
$3.35 best
1K documents20K tokens in, 1K out, each.
$250.00
$22.70 best
1K agent tasks200K tokens in, 20K out, each.
$3,000.00
$254.00 best
ContextMaximum tokens in one request.
1M
1M
AgenticMulti-step tasks with tools.
71.1 best
68.8
CodingCode writing and repair.
77.7 best
71.4
ReasoningLogic problems and puzzles.
75.8 best
67.4
MultimodalTasks with images and text.
77.9 best
62.2
KnowledgeFacts and expert knowledge.
77.2 best
63.5
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
74.3
—
MathMath problems.
77.0 best
61.7

Benchmarks

42 shared
Only Claude Fable 5.1: 90Only Step 5 Preview: 20
Claude Fable 5.1
Step 5 Preview
Artificial Analysis Humanity's Last ExamKnowledge
59.1% best3 / 222
46.5%23 / 222
Artificial Analysis Long Context ReasoningReasoning
85.3%5 / 217
88.3% best2 / 217
Critical Physics TasksReasoning
29.7% best12 / 214
20.9%26 / 214
Artificial Analysis Omniscience AccuracyKnowledge
68.4% best2 / 164
36.4%64 / 164
GDPval-AA normalizedAgentic
62.7% best6 / 133
54.3%24 / 133
Vibe Code Bench v1.1Coding
90.3% best6 / 123
70.4%48 / 123
Artificial Analysis SciCodeCoding
63.1% best3 / 114
58.9%13 / 114
Legal Research BenchLegal
55.3% best2 / 90
40.4%35 / 90
Tax Agent BenchFinance
50.7% best2 / 82
27.3%39 / 82
LiveBench Data AnalysisReasoning
80.3% best9 / 80
63.5%69 / 80
LiveBench ReasoningReasoning
91.7% best5 / 80
83.8%44 / 80
Finance Agent (v2)Finance
58.9% best8 / 79
50.7%42 / 79
Code MigrationCoding
54.6% best11 / 77
39.0%31 / 77
Humanity's Last ExamKnowledge
65.0% best3 / 77
46.5%26 / 77
EMBFinance
76.7% best1 / 75
57.1%41 / 75
Terminal-Bench 4.0Agentic
58.1% best6 / 61
31.8%23 / 61
DeepSWEAgentic
67.4%29 / 59
67.7% best27 / 59
Agent Arena steerabilityAgentic
14.8 best5 / 54
−0.930 / 54
Agent Arena task outcomeAgentic
14.4 best5 / 54
−1.535 / 54
ProofBench v1.1Math
100.0% best1 / 54
42.0%37 / 54