Compare / Qwen3.5 4B vs Qwen3 4B Thinking 2507
Qwen3.5 4B vs Qwen3 4B Thinking 2507
Updated November 2026
Verdict
Qwen3.5 4B scores higher on the Small LM Index: 57.8 vs 55.7, although their likely ranges overlap, so the lead is not certain. On the 3 benchmarks both labs report, Qwen3.5 4B wins 3 and Qwen3 4B Thinking 2507 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 Qwen3.5 4B if…
- you want the strongest model overall
- you mostly write code
- you need image input
- you want the more widely used model (more quantized builds and community support)
Pick Qwen3 4B Thinking 2507 if…
No clear advantage in the data we have.
Specifications
| Qwen3.5 4B | Qwen3 4B Thinking 2507 | |
|---|---|---|
| Small LM Index | 57.8 (likely 52–63) | 55.7 (likely 51–60) |
| Lab | Alibaba (Qwen) · China | Alibaba (Qwen) · China |
| Released | 2026-03-02 | 2025-08-06 |
| Parameters | 4B (dense) | 4B (dense) |
| Memory at 4-bit | ~3.3 GB | ~4.2 GB |
| Context window | 262K tokens | 262K tokens |
| Input | text, image, video | text |
| Reasoning | Hybrid (thinking on/off) | Always thinks |
| License | Apache-2.0 | Apache-2.0 |
| API price / 1M tokens | Self-host only | Self-host only |
| Hugging Face downloads (30 days) | 6.1M | 440.9K |
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 | Qwen3.5 4B | Qwen3 4B Thinking 2507 |
|---|---|---|
| MMLU-Pro Knowledge | 79.1 | 74.0 |
| GPQA Diamond Knowledge | 76.2 | 65.8 |
| AIME Math | ≈ 81 | 81.3 |
| LiveCodeBench Coding | 55.8 | 55.2 |
| τ²-Bench Agentic | 79.9 | — |
| IFBench Instruction following | 59.2 | — |
| MMMU-Pro Vision | 65.4✓ | — |
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, Qwen3.5 4B or Qwen3 4B Thinking 2507?
Qwen3.5 4B scores higher on the Small LM Index: 57.8 vs 55.7, although their likely ranges overlap, so the lead is not certain. On the 3 benchmarks both labs report, Qwen3.5 4B wins 3 and Qwen3 4B Thinking 2507 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?
Both have a 262K-token context window.
Can I use Qwen3.5 4B and Qwen3 4B Thinking 2507 commercially?
Both are released under the Apache-2.0 license. Apache 2.0 and MIT allow commercial use; other licenses can add conditions, so read the license before shipping a product.