Model comparison

DeepSeek-V3.2-Speciale vs Qwen3 Max

Qwen3 Max is the stronger model overall, scoring 43.7 to 39.7 on the Noometry Index. DeepSeek-V3.2-Speciale costs 2.8× less per token, which makes it the better buy when Qwen3 Max's lead doesn't matter for your workload.

Last verified October 11, 2026. 0 shared benchmarks.

DeepSeek-V3.2-Speciale

DeepSeek

39.7

Rank #162 Reported

Qwen3 Max

Alibaba (Qwen)

43.7

Rank #87 Confirmed

Summary

The widest gap is in writing & preference, where Qwen3 Max leads 62.4 to 46.0.
DeepSeek-V3.2-Speciale is cheaper at $0.58 / $1.68 per million input/output tokens, against $1.20 / $6 for Qwen3 Max.
Qwen3 Max accepts more context: 262K tokens versus 128K.
DeepSeek-V3.2-Speciale has downloadable open weights; the other is API-only.

Side by side

DeepSeek-V3.2-Speciale and Qwen3 Max specifications
	DeepSeek-V3.2-Speciale	Qwen3 Max
Provider	DeepSeek	Alibaba (Qwen)
Noometry Index	39.7	43.7
Released	2025-12-01	2025-09-23
Weights	Open	Proprietary
Context window	128K	262K
Max output	128K	66K
Input $ / M tokens	$0.58	$1.20
Output $ / M tokens	$1.68	$6
Results tracked	3	33

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

Category by category

Coding Qwen3 Max leads

DeepSeek-V3.2-Speciale: 40.4 (#140), Qwen3 Max: 43.0 (#93)

Coding benchmarks
Benchmark	DeepSeek-V3.2-Speciale	Qwen3 Max
WeirdML	46.7%	—
LMArena Coding	—	1456
ALE-Bench	—	370.45

Agentic & Tool Use Not comparable

DeepSeek-V3.2-Speciale: —, Qwen3 Max: —

Agentic & Tool Use benchmarks
Benchmark	DeepSeek-V3.2-Speciale	Qwen3 Max
Vending-Bench 2	—	71.56

Reasoning DeepSeek-V3.2-Speciale leads

DeepSeek-V3.2-Speciale: 32.9 (#73), Qwen3 Max: 22.6 (#190)

Reasoning benchmarks
Benchmark	DeepSeek-V3.2-Speciale	Qwen3 Max
SimpleBench	52.6%	—
Kagi LLM Benchmark	—	72.5%
NYT Connections (extended)	—	30.1%
Chess Puzzles	—	4%
LMArena Hard Prompts	—	1448
Mystery Game Puzzles	—	5%
DTBench	—	82.1%
LMCA	—	28.3%
Epoch Capabilities Index	—	142.38

Math Not comparable

DeepSeek-V3.2-Speciale: —, Qwen3 Max: 38.7 (#131)

Math benchmarks
Benchmark	DeepSeek-V3.2-Speciale	Qwen3 Max
FrontierMath (Tiers 1-3)	—	18.9%
OTIS Mock AIME 2024-2025	—	73.3%
LMArena Math	—	1446
MATH Level 5	—	97.1%

Knowledge Not comparable

DeepSeek-V3.2-Speciale: —, Qwen3 Max: 48.1 (#78)

Knowledge benchmarks
Benchmark	DeepSeek-V3.2-Speciale	Qwen3 Max
GPQA Diamond	—	72.6%
SimpleQA Verified	—	48.7%
LMArena Expert	—	1455

Multilingual Not comparable

DeepSeek-V3.2-Speciale: —, Qwen3 Max: 53.7 (#62)

Multilingual benchmarks
Benchmark	DeepSeek-V3.2-Speciale	Qwen3 Max
LMArena Non-English	—	1429
LMArena Chinese	—	1478
LMArena French	—	1449
LMArena German	—	1463
LMArena Japanese	—	1397
LMArena Korean	—	1399
LMArena Russian	—	1428
LMArena Spanish	—	1462

Instruction Following Not comparable

DeepSeek-V3.2-Speciale: —, Qwen3 Max: 74.8 (#87)

Instruction Following benchmarks
Benchmark	DeepSeek-V3.2-Speciale	Qwen3 Max
LMArena Instruction Following	—	1419

Long Context Not comparable

DeepSeek-V3.2-Speciale: —, Qwen3 Max: 41.6 (#134)

Long Context benchmarks
Benchmark	DeepSeek-V3.2-Speciale	Qwen3 Max
Fiction.LiveBench	—	66.7%
CL-bench	—	14.5%
LMArena Longer Query	—	1438

Writing & Preference Qwen3 Max leads

DeepSeek-V3.2-Speciale: 46.0 (#222), Qwen3 Max: 62.4 (#76)

Writing & Preference benchmarks
Benchmark	DeepSeek-V3.2-Speciale	Qwen3 Max
LMArena Text	—	1439
LMArena Creative Writing	—	1402
EQ-Bench Creative Writing	1276	—
LMArena Multi-Turn	—	1446

Frequently asked questions

Is DeepSeek-V3.2-Speciale better than Qwen3 Max?

Which is cheaper, DeepSeek-V3.2-Speciale or Qwen3 Max?

DeepSeek-V3.2-Speciale is cheaper. It lists at $0.58 per million input tokens and $1.68 per million output tokens; Qwen3 Max lists at $1.20 and $6.

Is DeepSeek-V3.2-Speciale or Qwen3 Max better for coding?

Qwen3 Max scores higher on coding benchmarks: 43.0 versus 40.4 in the Noometry coding category.

Which has the bigger context window?

Qwen3 Max does, with 262K tokens against 128K.

How many benchmarks do DeepSeek-V3.2-Speciale and Qwen3 Max share?

0 benchmarks have published results for both models. DeepSeek-V3.2-Speciale has 3 scored results on Noometry and Qwen3 Max has 33.