Model Details
Qwen2.5 Coder 7B Instruct
Quick answer: Qwen2.5 Coder 7B Instruct is Alibaba's October 2024 code-specialised 7B model, scoring 88.4% HumanEval, 83.9% GSM8K, 49.6% ACEBench, and 20.3% BigCodeBench. Apache 2.0 — the compact Qwen coding model.
At a Glance
Where Qwen2.5 Coder 7B Instruct leads
- 88.4% HumanEval — strong coding at 7B size
- 76.8% HellaSwag — good language understanding
- Apache 2.0 — fully open
- Code-first training on code datasets
Where it lags
- 20.3% BigCodeBench — struggles with complex code generation
- 49.6% ACEBench — moderate agent/tool calling
- 83.9% GSM8K — slightly below Qwen2.5 7B Instruct (91.6%)
- 7B parameters: limited vs larger coder models
Best for: Compact code assistance and code completion; local/edge IDE plugins; cost-efficient coding agents; lightweight code review tools.
What Qwen2.5 Coder 7B Instruct Is
Qwen2.5 Coder 7B Instruct is the code-specialised variant in the Qwen2.5 Coder series, released October 2024. Unlike Qwen2.5 7B Instruct (general instruction), the Coder variant is trained on an extended code corpus with specific code instruction fine-tuning.
The code-specific training shows in HumanEval (88.4% vs 84.8% for 7B Instruct) but trades off some math performance (83.9% GSM8K vs 91.6%). This is the standard tradeoff for code-specialised models.
Specifications
| Field | Value |
|---|---|
| Organization | Alibaba |
| License | Apache 2.0 |
| HuggingFace | Qwen/Qwen2.5-Coder-7B-Instruct |
| Release date | October 7, 2024 |
| Parameters | 7B |
| Modality | Text / Code |
| Context window | 128K tokens |
Pricing
Open weights under Apache 2.0 — self-host at no cost. Available via major inference providers.
Public Benchmark Scores
| Benchmark | Score | Source | Date |
|---|---|---|---|
| HumanEval | 88.4% | Benchgen evaluation | 2024-10 |
| GSM8K | 83.9% | Benchgen evaluation | 2024-10 |
| HellaSwag | 76.8% | Benchgen evaluation | 2024-10 |
| ACEBench | 49.6% | Benchgen evaluation | 2024-10 |
| BigCodeBench | 20.3% | Benchgen evaluation | 2024-10 |
Qwen2.5 Coder 7B vs Alternatives
| Model | HumanEval | GSM8K | License | Focus |
|---|---|---|---|---|
| Qwen2.5 Coder 7B Instruct | 88.4% | 83.9% | Apache 2.0 | Code |
| Qwen2.5 7B Instruct | 84.8% | 91.6% | Apache 2.0 | General |
| Granite 3.3 8B Instruct | 89.7% | 80.9% | Apache 2.0 | General |
| Llama 3.1 8B Instruct | — | 84.5% | Llama 3.1 | General |
Qwen2.5 Coder 7B vs Qwen2.5 7B Instruct: higher HumanEval (88.4% vs 84.8%) but lower GSM8K (83.9% vs 91.6%). If coding is primary: Coder 7B. If balanced math + code: 7B Instruct.
Related Models & Benchmarks
- Compare: Qwen2.5 7B Instruct · Granite 3.3 8B Instruct · Llama 3.1 8B Instruct
- Benchmarks: HumanEval · BigCodeBench · ACEBench
Frequently Asked Questions
What is Qwen2.5 Coder 7B Instruct?
Alibaba's October 2024 code-specialised 7B model scoring 88.4% HumanEval, 83.9% GSM8K, and 20.3% BigCodeBench. Apache 2.0.What is the difference between Qwen2.5 7B Instruct and Qwen2.5 Coder 7B Instruct?
Both are 7B Apache 2.0 models. Coder 7B is trained on extended code data with code-specific fine-tuning (higher HumanEval: 88.4% vs 84.8%). 7B Instruct focuses on general instruction-following with better math (91.6% vs 83.9% GSM8K).Specs from Alibaba's Qwen2.5 Coder 7B Instruct release (October 2024) and Benchgen evaluations. Last updated 2026-07-24.
Benchmark Leaderboards
This model isn’t on any benchmark leaderboard yet.