Inference
EN

Claude Fable 5.1 vs Hy4

Change models
Claude Fable 5.1Anthropic
Hy4Tencent
IndexAll categories, out of 100.
76.2 ±2.4 best
72.0 ±7.8
GapIndex points below the top.
Top
−4.2 · no clear gap
CoverageIndex weight that has results.
95% best
65%
SpeedOutput tokens per second.
69/s best
52/s
EloLMArena rating and rank.
1512
—
ReleasedMonth of release.
2026-09
2026-11
Input / 1MUS dollars per 1M input tokens.
$10.00
$0.417 best
Output / 1MUS dollars per 1M output tokens.
$50.00
$1.25 best
Batch output / 1MBatch price per 1M output tokens.
$25.00
—
1K chats2K tokens in, 500 out, each.
$45.00
$1.46 best
1K documents20K tokens in, 1K out, each.
$250.00
$9.59 best
1K agent tasks200K tokens in, 20K out, each.
$3,000.00
$108.41 best
ContextMaximum tokens in one request.
1M
1.05M best
AgenticMulti-step tasks with tools.
71.1
74.5 best
CodingCode writing and repair.
77.7 best
68.4
ReasoningLogic problems and puzzles.
75.8
—
MultimodalTasks with images and text.
77.9 best
57.4
KnowledgeFacts and expert knowledge.
77.2
83.7 best
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
74.3
—
MathMath problems.
77.0
—

Benchmarks

13 shared
Only Claude Fable 5.1: 119Only Hy4: 8
Claude Fable 5.1
Hy4
Vibe Code Bench v1.1Coding
90.3% best6 / 123
81.0%30 / 123
Legal Research BenchLegal
55.3% best2 / 90
41.3%30 / 90
Tax Agent BenchFinance
50.7% best2 / 82
34.4%17 / 82
Humanity's Last ExamKnowledge
65.0% best3 / 77
59.1%7 / 77
Terminal-Bench 4.0Agentic
58.1% best6 / 61
39.4%15 / 61
DeepSWEAgentic
67.4%29 / 59
74.9% best8 / 59
Agent Arena steerabilityAgentic
14.8 best5 / 54
8.814 / 54
Agent Arena task outcomeAgentic
14.4 best5 / 54
6.712 / 54
AutomationBenchAgentic
31.4%26 / 32
56.2% best4 / 32
FrontierCode 2.0Coding
32.4% best8 / 27
31.0%9 / 27
ProgramBench: Can Language Models Rebuild Programs From Scratch?Coding
87.6% best3 / 19
30.0%12 / 19
OSWorld 2.1 (all 108 tasks)Agentic
80.7% best2 / 11
78.9%5 / 11
Terminal-Bench-Science 0.1Agentic
52.6% best8 / 11
19.8%11 / 11