BenchLM comparison · methodology bench-align-v5.8-2026-09-30

Claude Fable 5.2
Anthropic
93.65 /100
Estimated · public rank #1
vs
Gemini 4 Carbon
Google
89.15 /100
Supported · public rank #2
CategoryClaude Fable 5.2Gemini 4 CarbonReading
Agentic
Directional only
90.7
Estimated · #1/135#1/135
84.6
Estimated · #3/135#3/135
Directional only
Coding
Claude Fable 5.2 leads
89.2
Supported · #1/162#1/162
83.9
Supported · #2/162#2/162
Claude Fable 5.2 leads
Reasoning
Claude Fable 5.2 leads
91
Supported · #1/160#1/160
84.8
Supported · #2/160#2/160
Claude Fable 5.2 leads
Multimodal
Not comparable
90.4
Estimated · #1/67#1/67
—
  
Not comparable
Knowledge
Not comparable
98.5
Supported · #1/180#1/180
—
  
Not comparable
Multilingual
Not comparable
—
  
—
  
Not comparable
Instruction following
Not comparable
—
  
—
  
Not comparable
Math
Not comparable
93.8
Unranked · 2 rankable rowsUnranked
93.3
Unranked · 2 rankable rowsUnranked
Not comparable