AccelerateBench v1.1 - Speedup >= 2x: leaderboard

Share of 32 AccelerateBench v1.1 HPC program optimization tasks (arXiv:2611.20882v1, Tables 1-2) where the agent's final program passes correctness checks and achieves at least a 2x speedup (50% runtime reduction).

Metric: Tasks correct and >=2x faster (%). Source: arxiv.org. Saturation forecast: Around December 2027. 16 models tracked. Updated 21 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1Claude Sonnet 5.5High71.9#24
2Claude Opus 5.5High68.8#9
3GPT-6 AstraHigh53.1#10
4GPT-6.1 SolHigh50#15
5Hy4 previewHigh50#74
6hy4High46.9#40
7Kimi K3High40.6#61
8GLM-5.3High40.6#72
9GLM-5.2High37.5#150
10Grok 4.6High34.4#50
11DeepSeek V4 ProHigh34.4#196
12DeepSeek V4 FlashHigh34.4#267
13DeepSeek V4.1 Flash (1017)High34.4#47
14GLM-5.3 FlashHigh34.4#133
15Hy3High28.1#210

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=acceleratebench-v1-1-speedup-over-2x · How It Works · Data refreshed daily, snapshot 2026-11-21.