Compare / Qwen3 4B Thinking 2507 vs Phi-4-mini
Qwen3 4B Thinking 2507 vs Phi-4-mini
Updated November 2026
Verdict
Qwen3 4B Thinking 2507 scores higher on the Small LM Index: 59.7 vs 15.8. On the 8 benchmarks both labs report, Qwen3 4B Thinking 2507 wins 8 and Phi-4-mini wins 0. Both need about 4.2 GB of memory at 4-bit quantization. Neither model has a hosted API on OpenRouter, so you would run both yourself.
Pick Qwen3 4B Thinking 2507 if…
- you want the strongest model overall
- you mostly write code
- you need math and reasoning
- you need precise instruction following
- you build agents that call tools
- you work with long documents
Pick Phi-4-mini if…
No clear advantage in the data we have.
Specifications
| Qwen3 4B Thinking 2507 | Phi-4-mini | |
|---|---|---|
| Small LM Index | 59.7 (likely 57–62) | 15.8 (likely 15–17) |
| Lab | Alibaba (Qwen) · China | Microsoft (Phi) · United States |
| Released | 2025-08-06 | 2025-02-26 |
| Parameters | 4B (dense) | 3.8B (dense) |
| Memory at 4-bit | ~4.2 GB | ~4 GB |
| Context window | 262K tokens | 128K tokens |
| Input | text | text |
| Reasoning | Always thinks | No thinking |
| License | Apache-2.0 | MIT |
| API price / 1M tokens | Self-host only | Self-host only |
| Hugging Face downloads (30 days) | 437.1K | 340.5K |
Benchmarks
Higher is better. Bold marks the winner where both labs report a score. ≈ marks the index model's estimate for a score that wasn't reported, ✓ an independent measurement and * a score measured by another lab.
| Benchmark | Qwen3 4B Thinking 2507 | Phi-4-mini |
|---|---|---|
| MMLU-Pro Knowledge | 74.3✓ | 46.5✓ |
| GPQA Diamond Knowledge | 66.7✓ | 33.1✓ |
| Humanity's Last Exam Reasoning | 6.2✓ | 4.4✓ |
| AIME Math | 82.7✓ | 3.0✓ |
| LiveCodeBench Coding | 64.1✓ | 12.6✓ |
| τ²-Bench Agentic | 25.4✓ | 8.2✓ |
| IFBench Instruction following | 49.8✓ | 21.1✓ |
| MATH-500 Math | — | 69.6✓ |
| AA-LCR Long context | 37.3✓ | 15.3✓ |
Scores are mostly reported by each lab with its own settings, so small gaps are noise. See the methodology. Open in the interactive comparison to add more models.
Questions
Which is better, Qwen3 4B Thinking 2507 or Phi-4-mini?
Qwen3 4B Thinking 2507 scores higher on the Small LM Index: 59.7 vs 15.8. On the 8 benchmarks both labs report, Qwen3 4B Thinking 2507 wins 8 and Phi-4-mini wins 0.
Which is cheaper to run?
Neither model has a hosted API on OpenRouter, so you would run both yourself. Both need about 4.2 GB of memory at 4-bit quantization.
Which has the longer context window?
Qwen3 4B Thinking 2507 has the longer context window: 262K tokens against 128K.
Can I use Qwen3 4B Thinking 2507 and Phi-4-mini commercially?
Qwen3 4B Thinking 2507 is released under Apache-2.0 and Phi-4-mini under MIT. Apache 2.0 and MIT allow commercial use; other licenses can add conditions, so read the license before shipping a product.