Accelerate Bench
Coding agents speed up 32 real scientific applications, such as GROMACS and OpenFOAM, on a fixed CPU allocation within 8 hours. A hidden verifier checks correctness, and Score@1 averages the share of runtime saved, with incorrect programs scoring zero.
A correct 2x speedup earns 50% on a task. Paper results use version 1.1, where 16 agents scored 25.5% to 67.4%. 31 tasks are open on Harbor Hub, and the full 32 need approval on Hugging Face. Training on the tasks is prohibited.
- Made by
- Tencent Hy Team, HKUST, HKUST (Guangzhou), Tsinghua University and Peking University
- Released
- November 2026
- Kind
- Speed and optimization
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Code
- Tasks
- All public
- Measures
- Efficiency
- Language
- English
- Interaction
- Agentic
- Environment
- Terminal
- Items from
- Real work and use
- Refresh
- New versions replace it
- Run by
- Anyone
- In views
- Agents
- Strongest model tested
- GPT-6 Astra #4 on the capability index
- Last model added
- November 2026
- Scores refreshed
- November 20th, 2026
Claude Sonnet 5.5 leads with 67.4%
Best score per model from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best.
- 67.4 Claude Sonnet 5.5
- 59.4 Claude Opus 5.5
- 49.2 GPT-6.1 Sol
- 47.2 GPT-6 Astra
- 45.2
Tencent Hy4 preview
- 45.0
Tencent Hy4
- 42.5 Kimi K3
- 40.1
DeepSeek V4.1 Flash 1017
- 39.7 GLM-5.3
- 39.3 GLM-5.3-Flash
- 37.9
DeepSeek-V4-Pro
- 37.9 GPT-6.1 Terra
- 34.1
DeepSeek-V4-Flash
- 34.0 GLM-5.2
- 33.9 Grok 4.6
- 25.5
Tencent Hy3
- 1 Claude Sonnet 5.5 high effort; benchmark v1.1 67.4%
- 2 Claude Opus 5.5 high effort; benchmark v1.1 59.4%
- 3 GPT-6.1 Sol high effort; benchmark v1.1 49.2%
- 4 GPT-6 Astra high effort; benchmark v1.1 47.2%
- 5
Tencent Hy4 preview high effort; benchmark v1.1 45.2%
- 6
Tencent Hy4 high effort; benchmark v1.1 45%
- 7 Kimi K3 high effort; benchmark v1.1 42.5%
- 8
DeepSeek V4.1 Flash 1017 high effort; benchmark v1.1 40.1%
- 9 GLM-5.3 high effort; benchmark v1.1 39.7%
- 10 GLM-5.3-Flash high effort; benchmark v1.1 39.3%
- 11
DeepSeek-V4-Pro high effort; benchmark v1.1 37.9%
- 12 GPT-6.1 Terra high effort; benchmark v1.1 37.9%
- 13
DeepSeek-V4-Flash high effort; benchmark v1.1 34.1%
- 14 GLM-5.2 high effort; benchmark v1.1 34%
- 15 Grok 4.6 high effort; benchmark v1.1 33.9%
- 16
Tencent Hy3 high effort; benchmark v1.1 25.5%
Source: Official Accelerate Bench leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Sonnet 5.5 | 67.4% | high effort; benchmark v1.1, Claude Code | Sep 2026 | Official leaderboard |
| 2 | Claude Opus 5.5 | 59.4% | high effort; benchmark v1.1, Claude Code | Sep 2026 | Official leaderboard |
| 3 | GPT-6.1 Sol | 49.2% | high effort; benchmark v1.1, Codex | Sep 2026 | Official leaderboard |
| 4 | GPT-6 Astra | 47.2% | high effort; benchmark v1.1, Codex | Sep 2026 | Official leaderboard |
| 5 | | 45.2% | high effort; benchmark v1.1, Claude Code | Aug 2026 | Official leaderboard |
| 6 | | 45% | high effort; benchmark v1.1, Claude Code | Nov 2026 | Official leaderboard |
| 7 | Kimi K3 | 42.5% | high effort; benchmark v1.1, Claude Code | Jul 2026 | Official leaderboard |
| 8 | | 40.1% | high effort; benchmark v1.1, Claude Code | Oct 2026 | Official leaderboard |
| 9 | GLM-5.3 | 39.7% | high effort; benchmark v1.1, Claude Code | Aug 2026 | Official leaderboard |
| 10 | GLM-5.3-Flash | 39.3% | high effort; benchmark v1.1, Claude Code | Aug 2026 | Official leaderboard |
| 11 | | 37.9% | high effort; benchmark v1.1, Claude Code | Apr 2026 | Official leaderboard |
| 12 | GPT-6.1 Terra | 37.9% | high effort; benchmark v1.1, Codex | Oct 2026 | Official leaderboard |
| 13 | | 34.1% | high effort; benchmark v1.1, Claude Code | Apr 2026 | Official leaderboard |
| 14 | GLM-5.2 | 34% | high effort; benchmark v1.1, Claude Code | Jun 2026 | Official leaderboard |
| 15 | Grok 4.6 | 33.9% | high effort; benchmark v1.1, Claude Code | Aug 2026 | Official leaderboard |
| 16 | | 25.5% | high effort; benchmark v1.1, Claude Code | Jul 2026 | Official leaderboard |
Rankings by part
We read 32 parts from the benchmark’s own leaderboard in 8 groups: molecular simulation & electronic structure, fluids, atmosphere & ocean, transport, plasma & astronomy, bioinformatics & phylogenetics, numerics, tensors & graphs, quantum circuits, machine learning and scientific i/o. Each ranks models on one part of the test. Claude Opus 5.5 leads 15 of them, and 6 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
GROMACS
16 models we track have a score here.
- 1 Claude Sonnet 5.5 77.6%
- 2 Claude Opus 5.5 77.4%
- 3
Tencent Hy3 ↑13 57%
- 4 GPT-6 Astra 56.5%
- 5 GPT-6.1 Sol ↓2 54.4%
- 6
DeepSeek V4.1 Flash 1017 ↑2 54%
- 7
DeepSeek-V4-Flash ↑6 53.7%
- 8
Tencent Hy4 preview ↓3 53.2%
- 9
Tencent Hy4 ↓3 52.8%
- 10 GLM-5.3-Flash 52.4%
LAMMPS
15 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 85.3%
- 2
Tencent Hy4 ↑4 69.4%
- 3 GPT-6 Astra ↑1 68.3%
- 4 Claude Sonnet 5.5 ↓3 66.5%
- 5 GPT-6.1 Sol ↓2 60.7%
- 6
Tencent Hy3 ↑10 60.5%
- 7 GLM-5.2 ↑7 59.6%
- 8 GLM-5.3 ↑1 56.5%
- 9
Tencent Hy4 preview ↓4 51.8%
- 10
DeepSeek-V4-Pro ↑1 43.7%
GPAW
15 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 64.1%
- 2 Claude Sonnet 5.5 ↓1 57.9%
- 3
Tencent Hy4 ↑3 57.3%
- 4 GLM-5.3-Flash ↑6 48.6%
- 5 Kimi K3 ↑2 47.3%
- 6
Tencent Hy4 preview ↓1 43.5%
- 7 GPT-6 Astra ↓3 40.4%
- 8 GPT-6.1 Sol ↓5 38.1%
- 9 GPT-6.1 Terra ↑3 24.2%
- 10 GLM-5.3 ↓1 23.2%
Quantum ESPRESSO
13 models we track have a score here.
- 1 Claude Sonnet 5.5 73.4%
- 2 Claude Opus 5.5 61.9%
- 3
Tencent Hy4 preview ↑2 51.5%
- 4 GPT-6 Astra 46.7%
- 5 GPT-6.1 Terra ↑7 43.3%
- 6 GPT-6.1 Sol ↓3 39.2%
- 7
Tencent Hy4 ↓1 29.3%
- 8 GLM-5.3 ↑1 17.8%
- 9
DeepSeek-V4-Flash ↑4 16.3%
- 10 Grok 4.6 ↑5 11.6%
SIESTA
16 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 72.8%
- 2 Claude Sonnet 5.5 ↓1 69.2%
- 3
DeepSeek V4.1 Flash 1017 ↑5 54.5%
- 4 GLM-5.3 ↑5 53.3%
- 5 GPT-6 Astra ↓1 50.9%
- 6 GPT-6.1 Terra ↑6 49.2%
- 7 GPT-6.1 Sol ↓4 48.2%
- 8 GLM-5.3-Flash ↑2 42.6%
- 9 Kimi K3 ↓2 40.3%
- 10
Tencent Hy4 preview ↓5 39.4%
Fluidity
14 models we track have a score here.
- 1 Claude Sonnet 5.5 40.1%
- 2 Claude Opus 5.5 32.2%
- 3 Kimi K3 ↑4 29.2%
- 4
DeepSeek V4.1 Flash 1017 ↑4 27.2%
- 5 GLM-5.3-Flash ↑5 25.3%
- 6 GLM-5.3 ↑3 23.3%
- 7
DeepSeek-V4-Pro ↑4 19.9%
- 8 GPT-6.1 Sol ↓5 18.1%
- 9
DeepSeek-V4-Flash ↑4 17.6%
- 10 GLM-5.2 ↑4 16.5%
ICON
14 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 34%
- 2 Claude Sonnet 5.5 ↓1 33.2%
- 3 GPT-6 Astra ↑1 20.1%
- 4 GPT-6.1 Sol ↓1 15.5%
- 5 GLM-5.3 ↑4 8.10%
- 6 Grok 4.6 ↑9 5.40%
- 7 GLM-5.3-Flash ↑3 4.00%
- 8
Tencent Hy3 ↑8 3.60%
- 9
DeepSeek-V4-Flash ↑4 2.60%
- 10 GPT-6.1 Terra ↑2 1.30%
MPAS-A
12 models we track have a score here.
- 1 Claude Sonnet 5.5 47.1%
- 2
DeepSeek V4.1 Flash 1017 ↑6 31.1%
- 3
Tencent Hy4 preview ↑2 21.1%
- 4 GPT-6.1 Terra ↑8 19%
- 5 Claude Opus 5.5 ↓3 17.6%
- 6 GLM-5.3-Flash ↑4 16.1%
- 7 Kimi K3 15.4%
- 8
DeepSeek-V4-Flash ↑5 9.90%
- 9 GLM-5.2 ↑5 2.70%
- 10
Tencent Hy3 ↑6 0%
NEMO
16 models we track have a score here.
- 1 Claude Sonnet 5.5 59.5%
- 2
DeepSeek V4.1 Flash 1017 ↑6 31%
- 3
DeepSeek-V4-Pro ↑8 30%
- 4 Claude Opus 5.5 ↓2 23.4%
- 5 GLM-5.3-Flash ↑5 21.5%
- 6
Tencent Hy4 20.2%
- 7 GPT-6.1 Sol ↓4 12.9%
- 8
Tencent Hy3 ↑8 10.3%
- 9 Kimi K3 ↓2 10.2%
- 10 GPT-6 Astra ↓6 8.90%
OpenFOAM
14 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 85.8%
- 2 Claude Sonnet 5.5 ↓1 80.5%
- 3
Tencent Hy4 ↑3 75.1%
- 4
Tencent Hy4 preview ↑1 64.8%
- 5 GPT-6 Astra ↓1 58%
- 6 GPT-6.1 Sol ↓3 49.6%
- 7 Kimi K3 37.6%
- 8
DeepSeek-V4-Flash ↑5 27.8%
- 9 GPT-6.1 Terra ↑3 20%
- 10 GLM-5.2 ↑4 19.5%
Palabos
16 models we track have a score here.
- 1 Claude Sonnet 5.5 91.2%
- 2 Claude Opus 5.5 80.5%
- 3
DeepSeek-V4-Pro ↑8 79.6%
- 4
Tencent Hy4 preview ↑1 71.7%
- 5 Kimi K3 ↑2 64.3%
- 6 GPT-6.1 Sol ↓3 63.5%
- 7 GLM-5.2 ↑7 59.7%
- 8
DeepSeek-V4-Flash ↑5 58.8%
- 9 GLM-5.3-Flash ↑1 49.2%
- 10
DeepSeek V4.1 Flash 1017 ↓2 43.9%
SU2
14 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 79.3%
- 2 Claude Sonnet 5.5 ↓1 66.4%
- 3
Tencent Hy4 preview ↑2 56.3%
- 4 GPT-6 Astra 54.3%
- 5
Tencent Hy4 ↑1 46.2%
- 6 GPT-6.1 Terra ↑6 43.2%
- 7 GPT-6.1 Sol ↓4 41.8%
- 8 GLM-5.3 ↑1 41.7%
- 9
DeepSeek V4.1 Flash 1017 ↓1 41.7%
- 10
DeepSeek-V4-Flash ↑3 39.8%
OpenMC
16 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 55.2%
- 2 Claude Sonnet 5.5 ↓1 46%
- 3 GLM-5.3 ↑6 41.2%
- 4
DeepSeek V4.1 Flash 1017 ↑4 40.6%
- 5 GPT-6.1 Sol ↓2 39.6%
- 6 Kimi K3 ↑1 38.1%
- 7 GPT-6 Astra ↓3 37.8%
- 8 GLM-5.3-Flash ↑2 35.5%
- 9
Tencent Hy4 ↓3 33.1%
- 10
Tencent Hy4 preview ↓5 27.9%
ShengBTE
16 models we track have a score here.
- 1 Claude Sonnet 5.5 83.4%
- 2 Claude Opus 5.5 77.3%
- 3
DeepSeek V4.1 Flash 1017 ↑5 69.2%
- 4 GLM-5.3-Flash ↑6 66.2%
- 5
Tencent Hy4 preview 64.7%
- 6 GPT-6 Astra ↓2 63.7%
- 7 GLM-5.3 ↑2 63%
- 8
DeepSeek-V4-Pro ↑3 62.5%
- 9 Kimi K3 ↓2 62.3%
- 10 GLM-5.2 ↑4 56.6%
VPIC
13 models we track have a score here.
- 1 Claude Sonnet 5.5 49.5%
- 2 GPT-6.1 Terra ↑10 12.4%
- 3 GLM-5.3-Flash ↑7 11.2%
- 4
Tencent Hy4 preview ↑1 10.9%
- 5 GPT-6.1 Sol ↓2 3.90%
- 6
DeepSeek V4.1 Flash 1017 ↑2 2.70%
- 7
Tencent Hy4 ↓1 1.00%
- 8
DeepSeek-V4-Flash ↑5 0%
- 9
DeepSeek-V4-Pro ↑2 0%
- 10 GLM-5.2 ↑4 0%
PRESTO
14 models we track have a score here.
- 1 Claude Sonnet 5.5 47.6%
- 2
DeepSeek V4.1 Flash 1017 ↑6 32.9%
- 3
Tencent Hy4 ↑3 28.7%
- 4 GPT-6.1 Sol ↓1 25.8%
- 5 GPT-6.1 Terra ↑7 20.6%
- 6 Kimi K3 ↑1 20.2%
- 7
Tencent Hy4 preview ↓2 20.2%
- 8
DeepSeek-V4-Flash ↑5 19.5%
- 9 GLM-5.2 ↑5 17%
- 10 GPT-6 Astra ↓6 14.8%
ABySS
14 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 86.1%
- 2 Claude Sonnet 5.5 ↓1 85.5%
- 3 Kimi K3 ↑4 82.3%
- 4
DeepSeek V4.1 Flash 1017 ↑4 70.4%
- 5 GLM-5.3 ↑4 69.3%
- 6
Tencent Hy4 preview ↓1 68.8%
- 7 GPT-6.1 Sol ↓4 64.8%
- 8 GPT-6 Astra ↓4 62.6%
- 9 Grok 4.6 ↑6 62%
- 10
DeepSeek-V4-Flash ↑3 61.5%
MHM2
15 models we track have a score here.
- 1 Claude Sonnet 5.5 70.6%
- 2 Claude Opus 5.5 51.1%
- 3 GPT-6.1 Sol 47.1%
- 4
Tencent Hy4 ↑2 45.1%
- 5 Kimi K3 ↑2 43.2%
- 6 GLM-5.3 ↑3 42.1%
- 7
Tencent Hy4 preview ↓2 42%
- 8
DeepSeek V4.1 Flash 1017 40.5%
- 9 GPT-6.1 Terra ↑3 38.3%
- 10 GPT-6 Astra ↓6 36.1%
MrBayes
16 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 88.9%
- 2 Claude Sonnet 5.5 ↓1 68.8%
- 3 GPT-6.1 Sol 67.9%
- 4 GPT-6.1 Terra ↑8 46.2%
- 5 GPT-6 Astra ↓1 38.2%
- 6 GLM-5.3 ↑3 37.5%
- 7
DeepSeek V4.1 Flash 1017 ↑1 37.1%
- 8 Kimi K3 ↓1 26.8%
- 9
DeepSeek-V4-Flash ↑4 21.3%
- 10
Tencent Hy4 preview ↓5 17.6%
wtdbg2
13 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 75.6%
- 2 Claude Sonnet 5.5 ↓1 69.8%
- 3
Tencent Hy4 ↑3 63.1%
- 4 Kimi K3 ↑3 57.8%
- 5 GPT-6.1 Sol ↓2 57.2%
- 6 GPT-6 Astra ↓2 56.9%
- 7
DeepSeek-V4-Pro ↑4 34.6%
- 8 GPT-6.1 Terra ↑4 30.6%
- 9 GLM-5.3-Flash ↑1 29.2%
- 10 GLM-5.2 ↑4 26.9%
HPCG-derived
16 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 98.9%
- 2 GLM-5.3 ↑7 98.6%
- 3 Grok 4.6 ↑12 98.5%
- 4 GPT-6 Astra 95.9%
- 5 GLM-5.2 ↑9 93.5%
- 6
Tencent Hy3 ↑10 93.2%
- 7 GPT-6.1 Sol ↓4 93.2%
- 8 Claude Sonnet 5.5 ↓7 92.9%
- 9 GLM-5.3-Flash ↑1 87.2%
- 10
Tencent Hy4 ↓4 86.1%
HPL-derived
15 models we track have a score here.
- 1 GLM-5.3 ↑8 70.1%
- 2 Grok 4.6 ↑13 55.1%
- 3 Kimi K3 ↑4 50.4%
- 4
DeepSeek-V4-Flash ↑9 47.8%
- 5
Tencent Hy4 ↑1 41.9%
- 6 Claude Opus 5.5 ↓4 26.2%
- 7 GPT-6.1 Sol ↓4 25.1%
- 8
DeepSeek-V4-Pro ↑3 24.3%
- 9 Claude Sonnet 5.5 ↓8 23.1%
- 10 GLM-5.3-Flash 21.7%
Graph sparsification
15 models we track have a score here.
- 1
DeepSeek V4.1 Flash 1017 ↑7 91.6%
- 2 Claude Sonnet 5.5 ↓1 88.4%
- 3 GPT-6.1 Sol 85.9%
- 4 GPT-6 Astra 83.8%
- 5 Kimi K3 ↑2 81.1%
- 6 GPT-6.1 Terra ↑6 80.2%
- 7 GLM-5.3 ↑2 79.4%
- 8
Tencent Hy4 ↓2 78.1%
- 9 GLM-5.3-Flash ↑1 77.1%
- 10
DeepSeek-V4-Pro ↑1 76.6%
miniVite
16 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 53.4%
- 2
Tencent Hy4 preview ↑3 44.4%
- 3 Claude Sonnet 5.5 ↓2 38.6%
- 4
Tencent Hy3 ↑12 20.1%
- 5 Grok 4.6 ↑10 19.3%
- 6 GLM-5.3-Flash ↑4 19%
- 7 GPT-6 Astra ↓3 18.3%
- 8 GPT-6.1 Sol ↓5 18.1%
- 9
Tencent Hy4 ↓3 14.6%
- 10
DeepSeek V4.1 Flash 1017 ↓2 12.8%
NUFFT imaging
9 models we track have a score here.
- 1 GPT-6 Astra ↑3 76.3%
- 2 GLM-5.2 ↑12 72.4%
- 3 Claude Sonnet 5.5 ↓2 67.7%
- 4 Grok 4.6 ↑11 62.5%
- 5 GPT-6.1 Sol ↓2 60.7%
- 6
DeepSeek-V4-Pro ↑5 60%
- 7
Tencent Hy3 ↑9 51.9%
- 8
Tencent Hy4 ↓2 50.6%
- 9 GPT-6.1 Terra ↑3 23.8%
Tucker reconstruction
16 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 91.3%
- 2 Kimi K3 ↑5 90.3%
- 3
Tencent Hy4 ↑3 88.6%
- 4
DeepSeek V4.1 Flash 1017 ↑4 88.2%
- 5 GLM-5.3 ↑4 87.7%
- 6 GLM-5.3-Flash ↑4 85%
- 7
Tencent Hy4 preview ↓2 82%
- 8 Claude Sonnet 5.5 ↓7 79.6%
- 9 Grok 4.6 ↑6 78.5%
- 10 GPT-6.1 Terra ↑2 76.3%
QuEST
15 models we track have a score here.
- 1
DeepSeek-V4-Pro ↑10 97.4%
- 2 Kimi K3 ↑5 93.3%
- 3 Claude Opus 5.5 ↓1 82%
- 4 GPT-6.1 Terra ↑8 81.8%
- 5 Claude Sonnet 5.5 ↓4 80.5%
- 6
Tencent Hy4 preview ↓1 80.3%
- 7 GPT-6.1 Sol ↓4 75.4%
- 8 Grok 4.6 ↑7 72.1%
- 9 GPT-6 Astra ↓5 72%
- 10 GLM-5.2 ↑4 70.2%
QiboTN
16 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 99.6%
- 2 Grok 4.6 ↑13 99.5%
- 3
DeepSeek-V4-Flash ↑10 99.4%
- 4 Kimi K3 ↑3 99.4%
- 5
DeepSeek V4.1 Flash 1017 ↑3 99.4%
- 6
Tencent Hy3 ↑10 99.3%
- 7 GLM-5.3-Flash ↑3 99.3%
- 8
Tencent Hy4 ↓2 99.2%
- 9
DeepSeek-V4-Pro ↑2 99.1%
- 10 GLM-5.2 ↑4 98.9%
Paged LM generation
16 models we track have a score here.
- 1 GPT-6 Astra ↑3 93.5%
- 2 GLM-5.3-Flash ↑8 93.2%
- 3 GPT-6.1 Sol 92.9%
- 4
DeepSeek V4.1 Flash 1017 ↑4 92.9%
- 5 Claude Sonnet 5.5 ↓4 92.6%
- 6 Kimi K3 ↑1 92.2%
- 7
DeepSeek-V4-Flash ↑6 92%
- 8 GPT-6.1 Terra ↑4 92%
- 9 GLM-5.2 ↑5 91.8%
- 10 Grok 4.6 ↑5 91.6%
ThunderSVM
16 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 91.7%
- 2 Claude Sonnet 5.5 ↓1 89.3%
- 3 Kimi K3 ↑4 79.9%
- 4
DeepSeek-V4-Flash ↑9 79.5%
- 5 GLM-5.3-Flash ↑5 77.2%
- 6
Tencent Hy3 ↑10 73%
- 7 Grok 4.6 ↑8 72.6%
- 8
Tencent Hy4 preview ↓3 72.4%
- 9
DeepSeek V4.1 Flash 1017 ↓1 69.9%
- 10 GPT-6 Astra ↓6 69.6%
DataLife
16 models we track have a score here.
- 1
Tencent Hy4 preview ↑4 86.7%
- 2
Tencent Hy4 ↑4 85.7%
- 3 GLM-5.3-Flash ↑7 85.3%
- 4
DeepSeek V4.1 Flash 1017 ↑4 85.3%
- 5 Claude Sonnet 5.5 ↓4 84.5%
- 6 Kimi K3 ↑1 84.3%
- 7 GPT-6.1 Sol ↓4 83.6%
- 8 GLM-5.2 ↑6 82.9%
- 9 GPT-6.1 Terra ↑3 81.9%
- 10
Tencent Hy3 ↑6 80.7%
IO500-derived
14 models we track have a score here.
- 1 GLM-5.3 ↑8 53.1%
- 2
Tencent Hy4 ↑4 44.8%
- 3 Claude Opus 5.5 ↓1 44.2%
- 4 Claude Sonnet 5.5 ↓3 41.9%
- 5
Tencent Hy4 preview 11.3%
- 6 GPT-6.1 Sol ↓3 6.40%
- 7 GPT-6 Astra ↓3 6.10%
- 8
DeepSeek-V4-Pro ↑3 5.80%
- 9 Grok 4.6 ↑6 3.50%
- 10 GLM-5.2 ↑4 3.20%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Accelerate Bench leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Accelerate Bench leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Coding
- SWE-bench Verified In use 500 human-validated GitHub issues from 12 Python repos. Agent writes a patch that must pass hidden tests. Claude Opus 4.7 83.5%
- LiveCodeBench v6 In use Competitive programming problems collected after model cutoffs from LeetCode, AtCoder, Codeforces. Scored pass@1 on hidden tests.
DeepSeek-V4-Pro 93.5%
- SWE-bench Multilingual In use 300 GitHub-issue tasks across nine programming languages in SWE-bench format. Patch graded by repository tests. Claude Fable 5.2 96.6%
- DeepSWE In use 113 original long-horizon engineering tasks across 91 repos and five languages, written from scratch. Hand-written verifiers accept alternative solutions. GPT-6 Astra 74.1%
- SciCode In use Scientist-curated research coding problems split into subproblems across natural sciences. Code checked against test cases.
Seed2.1 Pro 59.8%
- Codeforces rating In use Model solves recent Codeforces contest problems. An Elo-style rating is computed from simulated contest performance.
DeepSeek V4.1 Flash 3,471
Data audit
All 11 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 5 pass
- Pass Independent runs Scores from the official leaderboard (16 entries).
- Pass Models scored 16 models we track.
- Pass Newest score The newest is from November 2026, for Tencent Hy4.
- Pass Frontier models 5 of the top 10 have a score, led by GPT-6 Astra (#4).
- Pass Headroom The best is 67.4%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on acceleratebench.com: 16 rows for models we track at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.