deepseek-coder-1.3B-instruct: benchmark results
DeepSeek's 1.3B instruct code model, initialized from deepseek-coder-1.3b-base and pretrained on 2 trillion tokens (87% code, 13% English/Chinese text) with a 16K context window. Provider: DeepSeek. Released 2023-10-29. Access: Open.
Aggregate Rating 1196 ± 19, rank #1585 of 1656 rated models, from 33 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EffiBench - NET | 2.82 | Normalized Execution Time | 85.4 |
| EvoEval Concise | 60.98 | Pass@1 (%) | 62 |
| EvoEval Subtle | 53 | Pass@1 (%) | 59 |
| EffiBench - NMU | 1.91 | Normalized Memory Usage | 58.5 |
| EvalPlus (HumanEval+ & MBPP+) | 57.6 | Pass@1 avg (%) | 54 |
| EvoEval | 40.96 | Pass@1 (%) | 54 |
| EvoEval Verbose | 57.32 | Pass@1 (%) | 53 |
| EvoEval Creative | 25 | Pass@1 (%) | 51 |
| EvoEval Combine | 9 | Pass@1 (%) | 44 |
| EvoEval Difficult | 20 | Pass@1 (%) | 41 |
| C3-Bench (Controllable Code Completion) - Scale Control | 5 | Instruction-Following Rate (%; 909 SCC tasks: AST node-type | 30 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 44.02 | MC2 (%) (0-shot) | 25.7 |
Interactive version: theaggregate.ai/model?slug=deepseek-coder-1-3b-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.