Qwen3-14B vs Qwen3-235B-A22B vs Qwen3-32B

Best cell per row highlighted. Null means undisclosed — never counted as zero.

Alibaba Cloud
Qwen3-14BGA
7.8
AI Panel
Alibaba Cloud
Qwen3-235B-A22BGA
8.5
AI Panel
Alibaba Cloud
Qwen3-32BGA
8.4
AI Panel

Identity & lifecycle

ProviderAlibaba CloudAlibaba CloudAlibaba Cloud
Family / tierQwen3 · MediumQwen3 · LargeQwen3 · Large
StatusGAGAGA
Released2025-04-292025-04-292025-04-29
Knowledge cutoff2024-102024-102024-10

Architecture & context

Context window131K131K131K
Max output tokens33K33K33K
Input modalitiestexttexttext
Reasoning modeoptionaloptionaloptional
Open weightsYesYesYes

Pricing (per Mtok)

Input$0.06$0.2$0.08
Output$0.2$0.6$0.28
Cached input———
Batch input———
Free tierYesYesYes

Speed

Output speed (tok/s)—68—
Time to first token (s)—0.78—
Latency tierfastmediummedium

Trust & deployment

Trains on inputsNoNoNo
LicenseApache-2.0Apache-2.0Apache-2.0
CloudsGCPGCPGCP
Compliance———

AI Panel scoring

Unified score7.88.58.4
Decision Maker7.588.5
Domain Strategist7.58.58
Finance Lead9.59.59.5
Domain Practitioner999
Power User77.57.5
Skeptic7.577.5
Value score9.59.59.5

Benchmarks

MMLU-Pro—82.865.54
GPQA Diamond—70—
AIME 2025—81.5—
LiveCodeBench—70.7—
LMArena Elo—14231347
LMArena Coding Elo—14721407

Scores link to their sources. Missing cells mean the vendor hasn't disclosed a result — honesty over padding.