BenchLM comparison · methodology bench-align-v6.0-2026-11-19
| Category | Claude Fable 5.2 | Gemini 4 Neon | Reading |
|---|---|---|---|
Agentic Claude Fable 5.2 leads | 92.8 Supported · #2/182#2/182 | 91.9 Supported · #3/182#3/182 | Claude Fable 5.2 leads |
Coding Gemini 4 Neon leads | 92.4 Supported · #2/180#2/180 | 92.9 Supported · #1/180#1/180 | Gemini 4 Neon leads |
Reasoning Claude Fable 5.2 leads | 93.7 Supported · #1/161#1/161 | 88.9 Supported · #2/161#2/161 | Claude Fable 5.2 leads |
Multimodal Directional only | 87.2 Supported · #2/119#2/119 | 80.9 Estimated · #5/119#5/119 | Directional only |
Knowledge Directional only | 92.5 Estimated · #1/158#1/158 | 87.9 Estimated · #3/158#3/158 | Directional only |
Multilingual Not comparable | 87.8 Estimated · #1/88#1/88 | — | Not comparable |
Instruction following Directional only | 87.7 Supported · #1/143#1/143 | 86.9 Estimated · #2/143#2/143 | Directional only |
Math Directional only | 87.3 Estimated · #1/125#1/125 | 83.9 Estimated · #2/125#2/125 | Directional only |