Claude Fable 5.1 made me a really nice animated pelican
2026-09-02 07:57Models🔥 34.9 heat scoredevto #14
1sources
1days unfolding
34.9heat score
5mentions
SummaryAI generated
On September 1, 2026, Anthropic released the Claude Fable 5.1 model. This model achieved a score of 52.6% in the Terminal-Bench-Science 0.1 benchmark test. In a Pelican plotting test, differences in performance were significant across different reasoning modes: the low and medium modes skipped reasoning steps; the xhigh mode generated 36,767 output tokens and took 7 minutes and 51 seconds; the max mode finally produced an SVG image containing a blue hat, a fish basket, and a detailed bicycle layout, consuming 65,927 output tokens and taking 13 minutes and 54 seconds, with a cost of $3.30.
Anthropic released Claude Fable 5.1, which achieved a score of 52.6% in the Terminal-Bench-Science 0.1 benchmark. The author conducted Pelican plotting tests on the model and found that it skipped inference steps in low and medium reasoning modes, while only performing minimal planning in high mode. In xhigh mode, it generated 36,767 output tokens in 7 minutes and 51 seconds; in max mode, it produced SVG images containing blue hats, fish baskets, and detailed bicycle layouts, using a total of 65,927 output tokens and taking 13 minutes and 54 seconds, at a cost of $3.30.