reasoning benchmark
PostTrainBench
PostTrainBench evaluates a model's ability to autonomously post-train base models. Given pretrain-only base models, the agent must complete the full pipeline of data synthesis, training, evaluation, and iteration within a time budget, scored across downstream benchmarks such as AIME2025, BFCL, GPQA Main, GSM8K, and HumanEval.
Updated Aug 7, 2026
Published models5
Registry coverage5
MetricScore
EvidenceB
PostTrainBench leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | MI | 37.1% | 100.0% | 5 | C | |
| 2 | MA | 36.6% | 75.0% | 5 | C | |
| 3 | ZA | 34.3% | 50.0% | 5 | C | |
| 4 | BY | 18.3% | 25.0% | 5 | C | |
| 5 | BY | 16.5% | 0.0% | 5 | C |
Score distribution
Top published rows on the benchmark's original scale.
PostTrainBench highlights
The top published results on this benchmark's own scale.
What is PostTrainBench?
Definition and scoring fields from the benchmark registry.
PostTrainBench evaluates a model's ability to autonomously post-train base models. Given pretrain-only base models, the agent must complete the full pipeline of data synthesis, training, evaluation, and iteration within a time budget, scored across downstream benchmarks such as AIME2025, BFCL, GPQA Main, GSM8K, and HumanEval.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- PostTrainBench
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- posttrainbench|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about PostTrainBench.
Which model scores highest on PostTrainBench?
MiniMax M3 is currently ranked first with 37.1%.
What does PostTrainBench measure?
PostTrainBench evaluates a model's ability to autonomously post-train base models. Given pretrain-only base models, the agent must complete the full pipeline of data synthesis, training, evaluation, and iteration within a time budget, scored across downstream benchmarks such as AIME2025, BFCL, GPQA Main, GSM8K, and HumanEval.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
5 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.