Claude pelican comparison: Fable, Opus and Sonnet
Prompt: “Generate an SVG of a pelican riding a bicycle”. Five thinking-effort levels, shown from max to low.
Prices per 1M input/output tokens: Claude Fable 5.1 $10/$50; Claude Opus 5.5 $4/$20; Claude Opus 5 $5/$25; Claude Sonnet 5 $2/$10. Cached input: Fable 5.1 $0.25/M; Opus 5.5 $0.20/M.
Calculated costs use the supplied prices and logged token counts, including thinking tokens in the output total. Input is charged at the uncached rate because the source logs do not specify cached usage. Costs are rounded to two decimal places in cents. Fable runs: September 1, 2026; all other runs: September 22, 2026.
| Effort | Claude Fable 5.1 | Claude Opus 5.5 | Claude Opus 5 | Claude Sonnet 5 |
|---|---|---|---|---|
| max | 59,725 thinking tokens. |
No SVG returnedThe response is empty. All 128,000 output tokens were thinking tokens. 128,000 thinking tokens. |
8,976 thinking tokens. |
66,485 thinking tokens. |
| xhigh | 33,128 thinking tokens. |
8,821 thinking tokens. |
3,372 thinking tokens. |
0 thinking tokens. |
| high | 305 thinking tokens. |
329 thinking tokens. |
1,792 thinking tokens. |
32 thinking tokens. |
| medium | 0 thinking tokens. |
220 thinking tokens. |
586 thinking tokens. |
0 thinking tokens. |
| low | 0 thinking tokens. |
0 thinking tokens. |
22 thinking tokens. |
0 thinking tokens. |
Fable 5.1 thinking-token counts are inferred: logged output tokens minus the full response text counted by the Anthropic token-counting API, with a six-token message-overhead correction calibrated against the low and medium runs, which contain no reasoning blocks. The API provides estimates; these are not recorded usage breakdowns. Calculation details.