OpenCoder-8B-Instruct: benchmark results
Infly's OpenCoder-8B-Instruct, initialized from OpenCoder-8B-Base and pretrained from scratch on 2.5 trillion tokens (90% code), then supervised fine-tuned on over 4.5 million examples, for English and Chinese. Provider: Infly. Released 2024-11-07. Access: Open.
Aggregate Rating 1368 ± 29, rank #1338 of 1656 rated models, from 23 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EvalPlus (HumanEval+ & MBPP+) | 74.4 | Pass@1 avg (%) | 88.7 |
| BigCodeBench | 43.2 | Pass@1 (%) | 69.6 |
| EvalPlus | 74.4 | EvalPlus Avg. (imported) | 50 |
| MBPP+ | 71.4 | MBPP+ pass@1 (imported) | 50 |
| EuroEval Portuguese NLU - ScaLA PT | 7.5 | Matthews correlation coefficient (%) | 40 |
| FullStackBench (2024-12 bilingual) | 43.63 | Pass@1 (%), 3,374 problems: 1,687 English and 1,687 Chinese; | 38.5 |
| HumanEval+ | 77.4 | HumanEval+ pass@1 (imported) | 37 |
| EuroEval Italian NLU - ScaLA IT | 9.48 | Matthews correlation coefficient (%) | 35.6 |
| EuroEval Italian NLU - Sentipolc16 | 42.93 | Matthews correlation coefficient (%) | 35.3 |
| EuroEval Portuguese NLU - SST-2 PT | 67.12 | Matthews correlation coefficient (%) | 32 |
| EuroEval Dutch NLU - DBRD | 69.4 | Matthews correlation coefficient (%) | 27.2 |
| EuroEval Norwegian Common Sense Reasoning | 14.03 | Common Sense Reasoning Average Score (%) | 27.1 |
Interactive version: theaggregate.ai/model?slug=opencoder-8b-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.