Benchmark
AI4AI-Bench
Benchmark by AI4AI-Bench authors
Paper v1
0.174 normalized score
Kimi K3 / Claude Code effort aggregate
frontier lab
Moonshot AI
Only Kimi K3 in the original AI4AI-Bench study is included.
Published evidence
Benchmark
PostTrainBench
Benchmark by PostTrainBench team
PostTrainBench v1.1 · September 2026 results
31.96%
Kimi K3, 1M context / Claude Code
Benchmark
PostTrainBench
Benchmark by PostTrainBench team
PostTrainBench v1.2 · September 30 release
31.48%
Kimi K3, 1M context / Claude Code
Benchmark
TasteVal
Benchmark by P-Zero Research
October 2026 paper v1 · 20-model main cohort
0.45× ×
Kimi K3 (reasoning on) / TasteVal
Benchmark
TasteVal
Benchmark by P-Zero Research
October 2026 paper v1 · 20-model main cohort
0.89× ×
Kimi K3 (reasoning on) / TasteVal
Benchmark
TasteVal
Benchmark by P-Zero Research
October 2026 paper v1 · 20-model main cohort
0.09× ×
Kimi K2.5 (max, 96k output) / TasteVal
Benchmark
TasteVal
Benchmark by P-Zero Research
October 2026 paper v1 · 20-model main cohort
0.23× ×
Kimi K2.5 (max, 96k output) / TasteVal