Compare / Nanbeige4.2 3B vs Qwen3 4B Thinking 2507

Nanbeige4.2 3B vs Qwen3 4B Thinking 2507

Updated November 2026

Verdict

Nanbeige4.2 3B scores higher on the Small LM Index: 67.2 vs 59.8. On the 4 benchmarks both labs report, Nanbeige4.2 3B wins 4 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 Nanbeige4.2 3B if…

you want the strongest model overall
you mostly write code

Pick Qwen3 4B Thinking 2507 if…

you want the more widely used model (more quantized builds and community support)

Specifications

	Nanbeige4.2 3B	Qwen3 4B Thinking 2507
Small LM Index	67.2 (likely 64–70)	59.8 (likely 57–62)
Lab	Nanbeige (Kanzhun) · China	Alibaba (Qwen) · China
Released	2026-07-21	2025-08-06
Parameters	4.2B (dense)	4B (dense)
Memory at 4-bit	~3.8 GB	~4.2 GB
Context window	262K tokens	262K tokens
Input	text	text
Reasoning	Always thinks	Always thinks
License	Apache-2.0	Apache-2.0
API price / 1M tokens	Self-host only	Self-host only
Hugging Face downloads (30 days)	120.5K	437.1K

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	Nanbeige4.2 3B	Qwen3 4B Thinking 2507
MMLU-Pro Knowledge	≈ 84	74.3^✓
GPQA Diamond Knowledge	87.4	66.7^✓
Humanity's Last Exam Reasoning	17.8	6.2^✓
AIME Math	≈ 92	82.7^✓
LiveCodeBench Coding	72.5	64.1^✓
SWE-bench Verified Coding	63.6	≈ 52
Terminal-Bench Agentic	44.1	—
τ²-Bench Agentic	—	25.4^✓
IFBench Instruction following	—	49.8^✓
HMMT Math	82.8	—
AA-LCR Long context	58.7	37.3^✓

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, Nanbeige4.2 3B or Qwen3 4B Thinking 2507?

Nanbeige4.2 3B scores higher on the Small LM Index: 67.2 vs 59.8. On the 4 benchmarks both labs report, Nanbeige4.2 3B wins 4 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 Nanbeige4.2 3B 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.