CodeLlama-13B: benchmark results
Provider: Meta.
Aggregate Rating 1294 ± 36, rank #8819 of 10916 rated models, from 17 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| BizBench - CodeTAT-QA | 82.1 | Accuracy (%; 392 TAT-QA table questions answered by generati | 73.3 |
| BizBench | 56.8 | Mean accuracy (%) over the eight BizBench tasks (FinKnow, Fo | 57.1 |
| BizBench - CodeFinQA | 39.1 | Accuracy (%; 844 FinQA questions over report text and tables | 53.3 |
| BizBench - ConvFinQA Extract | 87 | Accuracy (%; 916 ConvFinQA questions answerable by extractin | 46.7 |
| BizBench - FormulaEval | 56 | Accuracy (%; 50 financial-formula functions to complete from | 46.7 |
| BizBench - SEC-Num | 61.3 | Accuracy (%; 2,000 SEC 10-K and 10-Q filing snippets, extrac | 46.7 |
| BizBench - FinCode | 9.9 | Accuracy (%; 121 CFA and CPA exam questions answered by gene | 43.3 |
| EvalPlus (HumanEval+ & MBPP+) | 45.5 | Pass@1 avg (%) | 39.5 |
| BigCode Models Leaderboard | 35.1 | HumanEval Python Pass@1 (%) | 39 |
| BizBench - TAT-QA Extract | 81.5 | Accuracy (%; 248 TAT-QA questions answerable by extracting a | 36.7 |
| AgentBoard | 25.8 | Progress Rate (imported) | 33.3 |
| Big Code Memorization - HumanEval pass@50 | 24.05 | HumanEval pass@50 (%) | 11.8 |
Interactive version: theaggregate.ai/model?slug=codellama-13b · How It Works · Data refreshed daily, snapshot 2026-11-08.