Model comparison

Qwen3-4B vs Qwen3.8 Max

Qwen3.8 Max is the stronger model overall, scoring 56.8 to 31.9 on the Noometry Index.

Last verified October 10, 2026. 3 shared benchmarks.

Qwen3-4B

Alibaba (Qwen)

31.9

Rank #264 Confirmed

Qwen3.8 Max

Alibaba (Qwen)

56.8

Rank #22 Confirmed

Summary

They share 3 benchmarks with published results for both. Qwen3-4B scores higher in 0 categories and Qwen3.8 Max in 4 categories; 4 gaps are clear of the uncertainty.
The widest gap is in math, where Qwen3.8 Max leads 73.2 to 29.7.
The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 52.2% for Qwen3-4B and 100% for Qwen3.8 Max.
Qwen3-4B has downloadable open weights; the other is API-only.

Side by side

Qwen3-4B and Qwen3.8 Max specifications
	Qwen3-4B	Qwen3.8 Max
Provider	Alibaba (Qwen)	Alibaba (Qwen)
Noometry Index	31.9	56.8
Released	2025-04-29	2026-08-02
Weights	Open	Proprietary
Context window	—	1M
Max output	—	131K
Input $ / M tokens	—	$2
Output $ / M tokens	—	$6
Results tracked	6	39

Sponsored placements are available on pages like this one. Advertise on Noometry

Category by category

Coding Not comparable

Qwen3-4B: —, Qwen3.8 Max: 53.5 (#29)

Coding benchmarks
Benchmark	Qwen3-4B	Qwen3.8 Max
DeepSWE	—	57.5%
LMArena WebDev	—	1674
FrontierSWE	—	17.8%
SciCode	—	53.2%
LMArena Coding	—	1502

Agentic & Tool Use Qwen3.8 Max leads

Qwen3-4B: 27.6 (#100), Qwen3.8 Max: 45.4 (#14)

Agentic & Tool Use benchmarks
Benchmark	Qwen3-4B	Qwen3.8 Max
APEX-Agents	—	63.3%
Berkeley Function Calling Leaderboard	35.7%	—
τ²-bench Banking	—	55.1%
GDP.pdf	—	23.2%

Reasoning Qwen3.8 Max leads

Qwen3-4B: 19.2 (#268), Qwen3.8 Max: 54.4 (#26)

Reasoning benchmarks
Benchmark	Qwen3-4B	Qwen3.8 Max
Chess Puzzles	4%	40%
NYT Connections (extended)	—	88.3%
CritPt	—	20%
LMArena Hard Prompts	—	1496
Mystery Game Puzzles	—	38%
DTBench	—	92%
LMCA	—	46.2%
Epoch Capabilities Index	—	156.41

Math Qwen3.8 Max leads

Qwen3-4B: 29.7 (#240), Qwen3.8 Max: 73.2 (#20)

Math benchmarks
Benchmark	Qwen3-4B	Qwen3.8 Max
OTIS Mock AIME 2024-2025	52.2%	100%
FrontierMath (Tiers 1-3)	—	74.7%
FrontierMath Tier 4	—	46.3%
MathArena Final-Answer Competitions	38.5%	—
ProofBench	—	58%
LMArena Math	—	1499

Knowledge Qwen3.8 Max leads

Qwen3-4B: 33.0 (#208), Qwen3.8 Max: 61.7 (#27)

Knowledge benchmarks
Benchmark	Qwen3-4B	Qwen3.8 Max
GPQA Diamond	52.3%	92.7%
SimpleQA Verified	—	47.3%
Vectara Hallucination Rate	5.7%	—
LMArena Expert	—	1507

Multimodal Not comparable

Qwen3-4B: —, Qwen3.8 Max: 37.2 (#75)

Multimodal benchmarks
Benchmark	Qwen3-4B	Qwen3.8 Max
LMArena Vision	—	1314
Furniture Assembly	—	20%

Multilingual Not comparable

Qwen3-4B: —, Qwen3.8 Max: 56.7 (#18)

Multilingual benchmarks
Benchmark	Qwen3-4B	Qwen3.8 Max
LMArena Non-English	—	1472
LMArena Chinese	—	1538
LMArena French	—	1503
LMArena German	—	1483
LMArena Japanese	—	1467
LMArena Korean	—	1461
LMArena Russian	—	1481
LMArena Spanish	—	1492

Instruction Following Not comparable

Qwen3-4B: —, Qwen3.8 Max: 77.6 (#17)

Instruction Following benchmarks
Benchmark	Qwen3-4B	Qwen3.8 Max
LMArena Instruction Following	—	1479

Long Context Not comparable

Qwen3-4B: —, Qwen3.8 Max: 45.6 (#31)

Long Context benchmarks
Benchmark	Qwen3-4B	Qwen3.8 Max
LMArena Longer Query	—	1489

Writing & Preference Not comparable

Qwen3-4B: —, Qwen3.8 Max: 67.1 (#30)

Writing & Preference benchmarks
Benchmark	Qwen3-4B	Qwen3.8 Max
LMArena Text	—	1483
LMArena Creative Writing	—	1479
LMArena Multi-Turn	—	1489

Frequently asked questions

Is Qwen3-4B better than Qwen3.8 Max?