Qwen2 72B vs Qwen2.5 72B: Benchmark Comparison
Detailed comparison of Qwen2 72B and Qwen2.5 72B covering benchmarks, pricing, context window, and compliance.
Key Specifications
| Specification | Qwen2 72B | Qwen2.5 72B |
|---|---|---|
| Vendor | alibaba | alibaba |
| Version | 2-72b | 2.5-72b |
| Release Date | 2024-06-07 | 2024-09-19 |
| Context Window | 131072 tokens | 131072 tokens |
| Input Modalities | text | text |
| Output Modalities | text | text |
| License | Qwen License | Qwen License |
| SOC2 | ✗ | ✗ |
| HIPAA | ✗ | ✗ |
| GDPR | ✗ | ✗ |
| ISO 27001 | ✗ | ✗ |
Benchmark Results
| Benchmark | Qwen2 72B | Qwen2.5 72B | Winner |
|---|---|---|---|
| ARC | 94.4 | — | Qwen2 72B |
| BBH | 83.4 | 82.4 | Qwen2 72B |
| GPQA | 53.6 | — | Qwen2 72B |
| GSM8K | 87 | 88.4 | Qwen2.5 72B |
| HUMANEVAL | 76.5 | 86.6 | Qwen2.5 72B |
| IFEVAL | 76.9 | — | Qwen2 72B |
| MATH | 71.9 | 83.1 | Qwen2.5 72B |
| MMLU | 80.2 | 86.1 | Qwen2.5 72B |
| MUSR | 68.9 | — | Qwen2 72B |
| WINOGRANDE | 82.5 | — | Qwen2 72B |
Pricing Comparison
| Tier (per Mtok) | Qwen2 72B | Qwen2.5 72B |
|---|---|---|
| Input | $0.5 | $0.5 |
| Output | $0.8 | $0.8 |
| Cache Read | $0 | $0 |
| Cache Write | $0 | $0 |
Qwen2 72B mot Qwen2.5 72B
Modellöversikt
Qwen2 72B and Qwen2.5 72B are both notable options in the AI model market. This page compares their benchmarks, pricing, and compliance.
Nyckelspecifikationer
| Leverantör | Releasedatum | Kontextfönster | Licens |
|---|---|---|---|
| Alibaba / Alibaba | 2024-06-07 / 2024-09-19 | 131K / 131K | Qwen License / Qwen License |
Benchmarkprestanda
| Benchmark | Qwen2 72B | Qwen2.5 72B | Vinnare |
|---|---|---|---|
| ARC | 94.4 | — | A |
| BBH (BIG-Bench Hard) | 83.4 | 82.4 | A |
| GPQA | 53.6 | — | A |
| GSM8K (Grade School Math 8K) | 87.0 | 88.4 | B |
| HumanEval | 76.5 | 86.6 | B |
| IFEval | 76.9 | — | A |
| MATH | 71.9 | 83.1 | B |
| MMLU (Massive Multitask Language Understanding) | 80.2 | 86.1 | B |
| MUSR | 68.9 | — | A |
| WinoGrande | 82.5 | — | A |
Prisjämförelse
| Inmatning | Utmatning | Cacheläsning | Cacheskrivning |
|---|---|---|---|
| — / — | — / — | — / — | — / — |
per miljon tokens — A / B
Styrkor & Svagheter
Qwen2 72B
- ✅ MMLU score 80.2, strong knowledge reasoning.
- ✅ GSM8K 87.0, robust math reasoning.
- ⚠️ 闭源专有模型，不支持自托管。
Qwen2.5 72B
- ✅ MMLU score 86.1, strong knowledge reasoning.
- ✅ HumanEval 86.6, excellent code generation.
- ✅ GSM8K 88.4, robust math reasoning.
- ⚠️ 闭源专有模型，不支持自托管。
Redaktörens åsikt
Qwen2 72B and Qwen2.5 72B each have their strengths. Choose based on workload (code, long context, vision), referencing the tables above.
Vanliga frågor
Which model is better for coding tasks?
Refer to the HumanEval benchmark table; the model with a higher score is better suited for coding tasks.
Which model is cheaper?
Refer to the pricing comparison table above; the model with lower input/output prices is more cost-effective.
Which has a longer context window?
Refer to the key specifications table; the model with a larger context window is better for long documents.
Referenser
Editor's Take
See Editor's Take section.