Benchmarks / Creative Reimplementation / Result
Claude Sonnet 5
Rank 10Final
Claude Sonnet 5 by Anthropic in the Creative Reimplementation Benchmark. Rank 10 of 13 in the order set by the original authors of the reference demo (members of Jumalauta).
01 / Result
Recording and run details
- Benchmark version
- v1.0
- Generation
- Single shot
- Reasoning effort
- High
- Harness
- Claude Code 2.1.283
- Run date
- 2026-09-26
- Run environment
- Reference container (see the protocol)
- Run time
- 15 min 06 s
- Tool calls
- 147
- Cost
- Not yet published
- Tokens
- Not yet published
- Run note
- The output is an offline renderer: its page draws a frame only when asked (renderFrame(t)) and has no playback loop or audio. The run ended while the model's own frame export was about a third done. The recording steps the model's renderFrame(t) at 30 fps, as the model's own capture script does; no model code was changed.
- Evaluators
- The original authors of the reference demo (members of Jumalauta)
- Recording length
- 2 min 55 s
- Recording published
- 2026-09-26
Evaluators' notes
The original authors gave the following reasons for their order. They are reported here as the authors' views, not as findings of this site. No other evaluation comments have been published.
- Rank 10
Claude Sonnet 5
According to the original authors, the Claude Sonnet 5 version goes off in directions of its own instead of following the original.
