CodeLlama-13B-Instruct-fp16: benchmark results
TheBloke's Transformers-format FP16 conversion of Meta's CodeLlama-13B-Instruct, converted from Meta's original weights. The Instruct and Python variants remain separate releases. Provider: TheBloke. Access: Open.
Aggregate Rating 1232 ± 16, rank #9396 of 10916 rated models, from 68 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMLU-by-task - College Physics | 32.35 | Accuracy (%) | 80.6 |
| MMLU-by-task - Abstract Algebra | 31 | Accuracy (%) | 60.2 |
| MMLU-by-task - College Chemistry | 36 | Accuracy (%) | 59.4 |
| MMLU-by-task - High School Computer Science | 49 | Accuracy (%) | 59.1 |
| MMLU-by-task - Moral Scenarios | 26.82 | Accuracy (%) | 56.5 |
| MMLU-by-task - High School Physics | 31.13 | Accuracy (%) | 56.1 |
| MMLU-by-task - Econometrics | 28.95 | Accuracy (%) | 55.2 |
| MMLU-by-task - TruthfulQA MC2 | 45.88 | Accuracy (%) | 55.2 |
| MMLU-by-task - TruthfulQA MC1 | 29.38 | Accuracy (%) | 54.6 |
| MMLU-by-task - Marketing | 70.51 | Accuracy (%) | 51.7 |
| MMLU-by-task - High School Statistics | 37.96 | Accuracy (%) | 48.7 |
| MMLU-by-task - Electrical Engineering | 42.07 | Accuracy (%) | 47.6 |
Interactive version: theaggregate.ai/model?slug=codellama-13b-instruct-fp16 · How It Works · Data refreshed daily, snapshot 2026-11-08.