Model comparison

DeepSeek-V3 vs Qwen3.5 397B-A17B

Qwen3.5 397B-A17B is the stronger model overall, scoring 45.7 to 39.2 on the Noometry Index. DeepSeek-V3 costs 3.3× less per token, which makes it the better buy when Qwen3.5 397B-A17B's lead doesn't matter for your workload.

Last verified October 10, 2026. 24 shared benchmarks.

DeepSeek-V3

DeepSeek

39.2

Rank #157 Confirmed

Qwen3.5 397B-A17B

Alibaba (Qwen)

45.7

Rank #60 Confirmed

Summary

They share 24 benchmarks with published results for both. DeepSeek-V3 scores higher in 1 category and Qwen3.5 397B-A17B in 7 categories; 7 gaps are clear of the uncertainty.
The widest gap is in knowledge, where Qwen3.5 397B-A17B leads 53.3 to 37.5.
The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 37.8% for DeepSeek-V3 and 88.9% for Qwen3.5 397B-A17B.
DeepSeek-V3 is cheaper at $0.24 / $0.90 per million input/output tokens, against $0.60 / $3.60 for Qwen3.5 397B-A17B.
Qwen3.5 397B-A17B accepts more context: 262K tokens versus 164K.

Side by side

DeepSeek-V3 and Qwen3.5 397B-A17B specifications
	DeepSeek-V3	Qwen3.5 397B-A17B
Provider	DeepSeek	Alibaba (Qwen)
Noometry Index	39.2	45.7
Released	2024-12-26	2026-02-01
Weights	Open	Open
Context window	164K	262K
Max output	164K	66K
Input $ / M tokens	$0.24	$0.60
Output $ / M tokens	$0.90	$3.60
Results tracked	60	36

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

Category by category

Coding Too close to call

DeepSeek-V3: 41.9 (#99), Qwen3.5 397B-A17B: 41.7 (#108)

Coding benchmarks
Benchmark	DeepSeek-V3	Qwen3.5 397B-A17B
LMArena Coding	1368	1464
Aider Polyglot	55.1%	—
LMArena WebDev	—	1400
SciCode	35.8%	—
WeirdML	36.1%	—
BigCodeBench Instruct	50%	—
LiveBench Coding	70.9%	—
BigCodeBench Complete	62.2%	—
HumanEval+	86.6%	—
MBPP+	73%	—

Agentic & Tool Use Not comparable

DeepSeek-V3: —, Qwen3.5 397B-A17B: 33.2 (#51)

Agentic & Tool Use benchmarks
Benchmark	DeepSeek-V3	Qwen3.5 397B-A17B
APEX-Agents	—	24.9%
τ²-bench Airline	—	81.5%
τ²-bench Banking	—	9.8%
τ²-bench Retail	—	84.4%
τ²-bench Telecom	—	97.8%
METR Time Horizons	49.6%	—

Reasoning Qwen3.5 397B-A17B leads

DeepSeek-V3: 20.0 (#226), Qwen3.5 397B-A17B: 33.6 (#62)

Reasoning benchmarks
Benchmark	DeepSeek-V3	Qwen3.5 397B-A17B
Kagi LLM Benchmark	52.3%	73.7%
LMArena Hard Prompts	1365	1448
DTBench	64.8%	87.5%
LMCA	15.5%	37.9%
Epoch Capabilities Index	135.95	146.64
SimpleBench	27.2%	—
NYT Connections (extended)	—	58.9%
CritPt	0%	—
Chess Puzzles	—	13%
Thematic Generalization	—	65.1%
LiveBench Reasoning	65.8%	—
Mystery Game Puzzles	—	18%
LiveBench Data Analysis	60.9%	—
BIG-Bench Hard	87.5%	—
ForecastBench	59.1	—
HellaSwag	88.9%	—
LiveBench	66.9%	—
PIQA	84.7%	—
WinoGrande	85.2%	—

Math Qwen3.5 397B-A17B leads

DeepSeek-V3: 30.4 (#208), Qwen3.5 397B-A17B: 44.9 (#64)

Math benchmarks
Benchmark	DeepSeek-V3	Qwen3.5 397B-A17B
OTIS Mock AIME 2024-2025	37.8%	88.9%
LMArena Math	1373	1455
FrontierMath (Tiers 1-3)	—	31.2%
Omni-MATH	40.3%	—
LiveBench Math	73.5%	—
MATH Level 5	75.5%	—
FrontierMath (Feb 2025 set)	1.7%	—

Knowledge Qwen3.5 397B-A17B leads

DeepSeek-V3: 37.5 (#147), Qwen3.5 397B-A17B: 53.3 (#53)

Knowledge benchmarks
Benchmark	DeepSeek-V3	Qwen3.5 397B-A17B
GPQA Diamond	67.6%	86.4%
LMArena Expert	1352	1461
MMLU-Pro	72.3%	—
Confabulations	26.1%	—
Vectara Hallucination Rate	6.1%	—
GPQA (HELM)	53.8%	—
ARC (AI2) Challenge	95.3%	—
MMLU	87.2%	—
TriviaQA	82.9%	—

Multimodal Not comparable

DeepSeek-V3: —, Qwen3.5 397B-A17B: 40.0 (#42)

Multimodal benchmarks
Benchmark	DeepSeek-V3	Qwen3.5 397B-A17B
LMArena Vision	—	1262

Multilingual Qwen3.5 397B-A17B leads

DeepSeek-V3: 48.8 (#138), Qwen3.5 397B-A17B: 54.0 (#56)

Multilingual benchmarks
Benchmark	DeepSeek-V3	Qwen3.5 397B-A17B
LMArena Non-English	1358	1431
LMArena Chinese	1390	1501
LMArena French	1386	1461
LMArena German	1374	1447
LMArena Japanese	1333	1427
LMArena Korean	1321	1382
LMArena Russian	1372	1429
LMArena Spanish	1358	1443

Instruction Following Qwen3.5 397B-A17B leads

DeepSeek-V3: 73.0 (#125), Qwen3.5 397B-A17B: 75.2 (#73)

Instruction Following benchmarks
Benchmark	DeepSeek-V3	Qwen3.5 397B-A17B
LMArena Instruction Following	1345	1424
LiveBench Instruction Following	81.5%	—
IFEval	83.2%	—

Long Context Qwen3.5 397B-A17B leads

DeepSeek-V3: 34.0 (#248), Qwen3.5 397B-A17B: 44.1 (#71)

Long Context benchmarks
Benchmark	DeepSeek-V3	Qwen3.5 397B-A17B
LMArena Longer Query	1353	1442
Fiction.LiveBench	50%	—

Writing & Preference Qwen3.5 397B-A17B leads

DeepSeek-V3: 57.6 (#124), Qwen3.5 397B-A17B: 62.5 (#74)

Writing & Preference benchmarks
Benchmark	DeepSeek-V3	Qwen3.5 397B-A17B
LMArena Text	1375	1438
LMArena Creative Writing	1364	1401
EQ-Bench Creative Writing	1472	1478
LMArena Multi-Turn	1389	1447
Short-Story Creative Writing	77%	—
WildBench	83%	—
LiveBench Language	49.1%	—

Frequently asked questions

Is DeepSeek-V3 better than Qwen3.5 397B-A17B?

Which is cheaper, DeepSeek-V3 or Qwen3.5 397B-A17B?

DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; Qwen3.5 397B-A17B lists at $0.60 and $3.60.

Is DeepSeek-V3 or Qwen3.5 397B-A17B better for coding?

They score almost the same on coding (41.9 vs 41.7); test both on your own repository before choosing.

Which has the bigger context window?

Qwen3.5 397B-A17B does, with 262K tokens against 164K.

How many benchmarks do DeepSeek-V3 and Qwen3.5 397B-A17B share?

24 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and Qwen3.5 397B-A17B has 36.