CodeLlama-70B-Instruct-hf: benchmark results
Provider: Meta. Released 2024-01-29. Access: Open.
Aggregate Rating 1429 ± 21, rank #1082 of 1640 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SocialBias-Bench | 28.34 | Code bias score (%; share of executable generated snippets w | 100 |
| SocialBias-Bench - Age | 10.5 | Code bias score for age (%; share of executable generated sn | 100 |
| CanItEdit (Lazy) | 37.52 | pass@1 (%; share of the 105 hand-written Python editing prob | 79.2 |
| CanItEdit (Descriptive) | 45.05 | pass@1 (%; share of the 105 hand-written Python editing prob | 75 |
| SocialBias-Bench - Education | 13.94 | Code bias score for education (%; share of executable genera | 66.7 |
| SocialBias-Bench - Employment Status | 17.49 | Code bias score for employment status (%; share of executabl | 66.7 |
| SocialBias-Bench - Gender | 10.9 | Code bias score for gender (%; share of executable generated | 66.7 |
| SocialBias-Bench - Race | 7.81 | Code bias score for race (%; share of executable generated s | 66.7 |
| Open LLM Leaderboard v1 - GSM8K | 46.25 | Accuracy (%) (5-shot) | 60 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 50.44 | MC2 (%) (0-shot) | 48.3 |
| Open LLM Leaderboard v1 - MMLU | 56.4 | Accuracy (%) (5-shot) | 39 |
| Open LLM Leaderboard v1 - WinoGrande | 74.51 | Accuracy (%) (5-shot) | 37 |
Interactive version: theaggregate.ai/model?slug=codellama-70b-instruct-hf · How It Works · Data refreshed daily, snapshot 2026-11-07.