TensorBench - Runtime: leaderboard
TensorBench (2026) tests coding agents (a published CLI harness with its model) on feature-addition and refactoring tasks in Scorch, a compiler-based tensor framework with dense and sparse tensors; a task passes when the patched repository passes its randomized regression tests and the tests the agent adds. This board is the pass rate in percent on the 28 runtime tasks, such as kernel caching, memory management and autotuning; higher is better.
Metric: Pass rate (%) on the 28 runtime tasks (caching and dispatch, memory management, autotuning, work stealing); 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 2030. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 (xHigh) | 60.7 | #49 (GPT-5.5) |
| 2 | Qwen 3 Coder 480B A35B Instruct | 42.9 | #643 |
| 3 | Claude Opus 4.7 (xHigh) | 42.9 | #79 (Claude Opus 4.7) |
| 4 | Gemini 3.1 Pro (Preview) | 39.3 | #94 |
| 5 | Claude Opus 4.6 | 39.3 | #102 |
| 6 | GPT-5.4 (xHigh) | 35.7 | #123 (GPT-5.4) |
| 7 | GPT-5.3 Codex (xHigh) | 25 | #116 (GPT-5.3 Codex) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=tensorbench-runtime · How It Works · Data refreshed daily, snapshot 2026-11-13.