Inference
EN
Vals AI

ProgramBench

Can language models rebuild programs from scratch?

ModelsModels with a result.64
Top resultBest result on this benchmark.19.0%Claude Fable 5.2
Top-3 spreadPoints from first to third.12.0 pts
UpdatedDate the source changed the results.9 Nov 2026

Result and cost

52 of 65 results

Results

65 results
#ModelResultThe score from the source.CostUS dollars to run one task.TimeTime to run one task.
01Claude Fable 5.219.0% ±2.8$46.822.1 h
02Claude Opus 5.518.5% ±2.8$42.662.4 h
03Claude Fable 5.17.0% ±1.8$58.522.4 h
04Claude Sonnet 5.56.5% ±1.7$22.622.1 h
05GPT-6 Astra5.5% ±1.6$11.5727 min
06GPT-6.1 Sol3.0% ±1.2$2.8770 min
06Claude Opus 53.0% ±1.2$60.292.9 h
08Gemini 4 Argon2.5% ±1.1$17.6671 min
08Claude Haiku 5.52.5% ±1.1$6.8373 min
08Muse Spark 1.3 Max2.5% ±1.1$54.295.5 h
11Claude Fable 52.0% ±1.0$75.682.6 h
11GPT-6 Sol2.0% ±1.0$11.6746 min
11Kimi K32.0% ±1.0$70.485.7 h
14GPT-5.6 Sol1.5% ±0.9$15.2929 min
14GLM-5.31.5% ±0.9$21.964.3 h
16Gemini 3.8 Flash1.0% ±0.7$10.9347 min
16Grok 4.61.0% ±0.7$16.764.5 h
16Claude Opus 4.81.0% ±0.7$31.271.5 h
19Muse Wonder Full0.5% ±0.5$28.656.1 h
19MiMo-V2.6-Pro0.5% ±0.5$0.7273.0 h
19Muse Spark 1.30.5% ±0.5$11.5669 min
19Grok 4.70.5% ±0.5$46.493.6 h
19GPT-5.50.5% ±0.5$6.9522 min
19GPT-5.6 Terra0.5% ±0.5$5.3834 min
19DeepSeek V4.1 Flash 10170.5% ±0.5$0.89762 min
19Inkling 2 Small0.5% ±0.5$8.461.8 h
19GPT-6 Luna0.5% ±0.5$0.18129 min
19Muse Spark 1.20.5% ±0.5$2.4727 min
19GPT-5.4high effort0.5% ±0.5$2.2716 min
19MiMo-V2.6-Flash0.5% ±0.5$0.4713.4 h
19GLM-5.20.5% ±0.5$12.582.1 h
19Claude Sonnet 4.60.5% ±0.5—72 min
19Inkling-Small0.5% ±0.5$0.47929 min
34Gemini 3.7 Flash0.0%$7.1322 min
34Hy4 preview0.0%$12.812.2 h
34Qwen3.8 Max0.0%$12.365.9 h
34DeepSeek V4 Pro 08130.0%$3.4183 min
34GPT-5.6 Luna0.0%$0.94342 min
34GPT-5.40.0%—40 min
34Muse Spark 1.10.0%$13.065.7 h
34Claude Sonnet 50.0%$24.361.5 h
34Gemini 3.5 Flash0.0%—18 min
34Grok 4.50.0%$4.1219 min
34Gemini 3.6 Flash0.0%$5.9248 min
34GLM-5.3-Flash0.0%$1.605.5 h
34Claude Opus 4.70.0%—54 min
34DeepSeek V4 Flash 07310.0%$1.295.6 h
34Gemini 3.1 Pro Preview0.0%$2.6313 min
34Qwen3.8-27B0.0%$17.995.9 h
34Kimi K2.60.0%—80 min
34Kimi K2.7 Code0.0%$3.1964 min
34DeepSeek V4 Pro 04230.0%—63 min
34Gemini 3 Flash Preview0.0%$0.447 min
34GLM-5.10.0%—66 min
34Inkling0.0%$4.1446 min
34Qwen3.6 Plus0.0%—33 min
34GPT-5.4 mini0.0%—54 min
34Grok 4.30.0%$0.3564 min
34Gemini 3.5 Flash-Lite0.0%$0.3377 min
34MiniMax M2.70.0%—24 min
34Gemini 3.1 Flash Lite Preview0.0%$0.0643 min
34Claude Haiku 4.5 Thinking0.0%—8 min
34Laguna M.10.0%—82 min
34Laguna XS.20.0%—26 min
34Nemotron Lightning 3p5 30B A3B0.0%$0.21724 min