TensorBench - Sparse Formats: leaderboard

Metric: Pass rate (%) on the 20 tasks that add new sparse tensor formats; each coding agent (a published CLI scaffold with its model) edits the open-source Scorch compiler-based tensor framework inside a per-task container, one run per task; a task passes when every test of the post-patch suite (pre-existing randomized regression tests plus the tests the agent adds) succeeds; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 7 models tracked.

Top models

#ModelScore
1Claude Opus 4.7 (xHigh)95
2GPT-5.5 (xHigh)55
3GPT-5.4 (xHigh)50
4Claude Opus 4.640
5GPT-5.3 Codex (xHigh)35
6Gemini 3.1 Pro (Preview)30
7Qwen 3 Coder 480B A35B Instruct15

Interactive version: theaggregate.ai/benchmark?slug=tensorbench-sparse-formats · How It Works · Data refreshed daily, snapshot 2026-09-29.