Phi-3 Medium vs Phi-3 Small: Benchmark Comparison
Detailed comparison of Phi-3 Medium and Phi-3 Small covering benchmarks, pricing, context window, and compliance.
Key Specifications
| Specification | Phi-3 Medium | Phi-3 Small |
|---|---|---|
| Vendor | other | other |
| Version | phi-3-medium | phi-3-small |
| Release Date | 2024-05-21 | 2024-05-21 |
| Context Window | 4096 tokens | 8192 tokens |
| Input Modalities | text | text |
| Output Modalities | text | text |
| License | MIT | MIT |
| SOC2 | ✗ | ✗ |
| HIPAA | ✗ | ✗ |
| GDPR | ✗ | ✗ |
| ISO 27001 | ✗ | ✗ |
Benchmark Results
| Benchmark | Phi-3 Medium | Phi-3 Small | Winner |
|---|---|---|---|
| ARC | 90.1 | 89.6 | Phi-3 Medium |
| BBH | 71.4 | 71.8 | Phi-3 Small |
| GPQA | 25.6 | 30.9 | Phi-3 Small |
| GSM8K | 59 | 62.7 | Phi-3 Small |
| HUMANEVAL | 53.9 | 68.3 | Phi-3 Small |
| IFEVAL | 67.4 | 61 | Phi-3 Medium |
| MATH | 34.3 | 43.5 | Phi-3 Small |
| MMLU | 62.1 | 71.8 | Phi-3 Small |
| MUSR | 43.9 | 36.5 | Phi-3 Medium |
| WINOGRANDE | 72.1 | 74.9 | Phi-3 Small |
Pricing Comparison
| Tier (per Mtok) | Phi-3 Medium | Phi-3 Small |
|---|---|---|
| Input | $0.2 | $0.15 |
| Output | $0.2 | $0.15 |
| Cache Read | $0 | $0 |
| Cache Write | $0 | $0 |
Phi-3 Medium gegen Phi-3 Small
Modellübersicht
Phi-3 Medium and Phi-3 Small are both notable options in the AI model market. This page compares their benchmarks, pricing, and compliance.
Hauptspezifikationen
| Anbieter | Veröffentlichungsdatum | Kontextfenster | Lizenz |
|---|---|---|---|
| Other / Other | 2024-05-21 / 2024-05-21 | 4K / 8K | MIT / MIT |
Benchmark-Leistung
| Benchmark | Phi-3 Medium | Phi-3 Small | Gewinner |
|---|---|---|---|
| ARC | 90.1 | 89.6 | A |
| BBH (BIG-Bench Hard) | 71.4 | 71.8 | Tie |
| GPQA | 25.6 | 30.9 | B |
| GSM8K (Grade School Math 8K) | 59.0 | 62.7 | B |
| HumanEval | 53.9 | 68.3 | B |
| IFEval | 67.4 | 61.0 | A |
| MATH | 34.3 | 43.5 | B |
| MMLU (Massive Multitask Language Understanding) | 62.1 | 71.8 | B |
| MUSR | 43.9 | 36.5 | A |
| WinoGrande | 72.1 | 74.9 | B |
Preisvergleich
| Eingabe | Ausgabe | Cache-Lesen | Cache-Schreiben |
|---|---|---|---|
| — / — | — / — | — / — | — / — |
pro Million Token — A / B
Stärken & Schwächen
Phi-3 Medium
- ✅ 可靠的通用模型。
- ⚠️ 闭源专有模型，不支持自托管。
- ⚠️ 上下文窗口 4K 偏小。
Phi-3 Small
- ✅ 可靠的通用模型。
- ⚠️ 闭源专有模型，不支持自托管。
- ⚠️ 上下文窗口 8K 偏小。
Redaktionsmeinung
Phi-3 Medium and Phi-3 Small each have their strengths. Choose based on workload (code, long context, vision), referencing the tables above.
FAQ
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.
Referenzen
Editor's Take
See Editor's Take section.