Claude Fable 5.2 vs Claude Fable 5.1: A Big Jump and Leads the Charts!
Anthropic has officially announced the latest version of Fable, its 5.2 model, a significant jump over its predecessor 5.1. Here is how both models compare with each other.
Benchmarks
| Benchmark | Claude Fable 5.1 | Claude Fable 5.2 | Change |
|---|---|---|---|
| FrontierCode 2.0 | 32.4%Xhigh effort | 49.1%Xhigh effort | +16.7 points |
| FrontierSWE v2 | 56.3% | 69.9% | +13.6 points |
| AutomationBench | 31.4% | 51.4% | +20.0 points |
| Terminal-Bench-Science 0.1 | 52.6% | 74.6% | +22.0 points |
| WANDR | 68.7% | 82.1% | +13.4 points |
| OSWorld 2.1 | 83.5% | 89.3% | +5.8 points |
| RoboDojo Sim | 27.6% | 64.9% | +37.3 points |
| AA-Briefcase v1.1 rubric | 61.5% | 66.1% | +4.6 points |
| AA-Briefcase v1.1 Elo | 1669 Elo | 1785 Elo | +116 Elo |
| Remote Labor Index | 17.9% | 41.3%High effort | +23.4 points |
| Humanity's Last Exam | 65.6% | 74.2% | +8.6 points |
| Chartography | 44.8% | 73.4% | +28.6 points |
Scores use each model's highest available effort unless a cell says otherwise.
Artificial Analysis benchmarks
| Benchmark | Fable 5.1, Max effort | Fable 5.2, Max effort | Change |
|---|---|---|---|
| Intelligence Index v5.0 | 49.2 | 64.6 | +15.4 |
| AA-Briefcase v1.1 | 1669 Elo | 1785 Elo | +116 Elo |
| AutomationBench-AA | 59.4% | 76.4% | +17.0 points |
| Terminal-Bench 4.0 | 52.0% | 59.6% | +7.6 points |
| FrontierCode 2.0 | 31.6% | 48.8% | +17.2 points |
| AA-Omniscience v1.1 | 46.3 | 66.8 | +20.5 |
| GDP.pdf | 26.2% | 24.2% | -2.0 points |
| HLE-Diamond | 60.1% | 80.2% | +20.1 points |
| CritPt v2 | 52.5% | 76.5% | +24.0 points |
| Terminal-Bench-Science 0.1 | 43.3% | 73.3% | +30.0 points |
Fable 5.2 wins 9 of these 10. Fable 5.1 keeps a 2-point lead on GDP.pdf.
Cost and tokens
| Metric | Fable 5.1, Max effort | Fable 5.2, Max effort | Change |
|---|---|---|---|
| Cost per Intelligence Index task | $12.99 | $26.54 | About 2x more |
| Output tokens per task | About 118K | About 274K | About 2.3x more |
| Reasoning tokens per task | About 71K | About 193K | About 2.7x more |
| Output tokens for the full Index | About 163M | About 275M | About 69% more |
| Output speed | 87 tokens/s | 103 tokens/s | About 18% faster |
Fable 5.2 is cheaper per token, but at Max effort it thinks far longer, so each Intelligence Index task ends up costing about twice as much.
Pricing and specs
| Spec | Claude Fable 5.1 | Claude Fable 5.2 | Change |
|---|---|---|---|
| Input | $10 | $8 | 20% cheaper |
| Output | $50 | $40 | 20% cheaper |
| Cache reads | $0.25 | $0.20 | 20% cheaper |
| Cache writes, 5 minutes | $12.50 | $10 | 20% cheaper |
| Cache writes, 1 hour | $20 | $16 | 20% cheaper |
| Batch input | $5 | $4 | 20% cheaper |
| Batch output | $25 | $20 | 20% cheaper |
| Fast mode | Not offered | $16 and $80Up to 2.5x faster generation | New |
| Context window | 1M tokens | 1M tokens | Same |
| Max output | 128K tokens | 128K tokens | Same |
| Thinking | AdaptiveAlways on | AdaptiveAlways on | Same |
| Default effort | High | High | Same |
| Input and output | Text and images in, text out | Text and images in, text out | Same |
| Knowledge cutoff | June 2026 | June 2026 | Same |
| Status | LegacyStill available | Latest | Fable 5.2 replaces 5.1 |
| Released | September 1, 2026 | November 9, 2026 | About 10 weeks apart |
| API model ID | claude-fable-5-1 | claude-fable-5-2 | New ID |