Claude Fable 5.2
Demanding reasoning. Hard reasoning and long-horizon agents.
Which reasoning effort setting?
controls how long Claude Fable 5.2 thinks before answering. Token prices never change; higher settings spend more and raise task cost.
Start at Max. The top setting is also the best value here. Rule: cheapest setting within 2 points of the model’s best score.
At Max, Fable 5.2 thinks for about 193K tokens per task, then writes about 81.1K more as its answer: $26.54 per task, 8.6× Low, for +10.7 points.
Output rate: $40.00/1M at every effort level. Thinking tokens are billed as output.
| Effort | Index | Per task | Thinking | All output | × Low | Time |
|---|---|---|---|---|---|---|
| Low | 53.9 | $3.10 | 16.5K | 35.9K | 1.00× | 5.5 min |
| Medium | 56.4 | $3.80 | 20.7K | 43.1K | 1.23× | 6.5 min |
| High | 61.8 | $9.08 | 58.2K | 97.8K | 2.93× | 15.1 min |
| Xhigh | 62.2 | $14.83 | 106K | 162K | 4.79× | 24.5 min |
| MaxStart here | 64.6 | $26.54 | 193K | 274K | 8.57× | 41.9 min |
Per task of the Artificial Analysis Intelligence Index: the average over its nine evaluations, counting input, cache, and output tokens at list prices. These are long, agent-style benchmark tasks, much longer than a chat message.
Where Claude Fable 5.2 sits on the frontier
Intelligence Index against one 100K-input, 10K-output request.