starcoder2-7B: benchmark results
BigCode's StarCoder2-7B, a 7B code model trained on 17 programming languages from The Stack v2 using Fill-in-the-Middle over 3.5+ trillion tokens, not an instruction-following model. Provider: BigCode. Released 2024-02-28. Access: Open.
Aggregate Rating 1214 ± 15, rank #1572 of 1656 rated models, from 51 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - GSM8K | 25.09 | Accuracy (%) (5-shot) | 43.9 |
| EvoEval Tool Use | 46 | Pass@1 (%) | 43 |
| EffiBench - NET | 3.02 | Normalized Execution Time | 42.7 |
| ShaderMatch | 10.15 | Clone Match Rate (%) | 38.1 |
| BigCode Models Leaderboard | 34.1 | HumanEval Python Pass@1 (%) | 35.6 |
| GSM8K | 32.7 | Accuracy (%) | 33.3 |
| CRUXEval | 36 | Output Prediction pass@1 (%) | 32.1 |
| Open LLM Leaderboard - MuSR | 37.93 | Score | 29.1 |
| EvoEval Creative | 17 | Pass@1 (%) | 24 |
| Big Code Memorization - HumanEval pass@1 | 36.59 | HumanEval pass@1 (%) | 23.5 |
| Big Code Memorization - HumanEval pass@50 | 29.05 | HumanEval pass@50 (%) | 23.5 |
| Big Code Memorization - HumanEval-ET pass@1 | 32.32 | HumanEval-ET pass@1 (%) | 23.5 |
Interactive version: theaggregate.ai/model?slug=starcoder2-7b · How It Works · Data refreshed daily, snapshot 2026-10-20.