Qwen3 30B A3B 2507 (Reasoning)
Jul 2025
Qwen3 30B A3B 2507 (Reasoning)Compare inon
Qwen3 30B A3B 2507 (Reasoning) measured on 13 of 104 Chat & Knowledge benchmarks13
BenchmarkByScoreRank
Overall1
Knowledge4
Science2
Documents2
Legacy · saturated or no longer reported4
No benchmark matches
$0.066per task?
price not verifiedprice
—output speed?
262Kcontext
Alternatives?
ModelScoreCostPer taskWhat changes
CheaperMiMo v2.5 ProthinkingOpen×3.5-18%$0.054Memory (4-bit): 18 GB → 614 GB
BetterGPT-6 Lunalow+97%-93%$0.005Function calling only with none · Chat Completions+3OpenQwen3.8 27Bxhigh×3.3+1,428%$1.01max_tokens range · QwenCloud+4CheaperMiMo v2.6 FlashthinkingOpen×3.4-5%$0.062Memory (4-bit): 18 GB → 186 GB+4
CheaperClaude Haiku 5.5medium×3.3-29%$0.047Remove sampling parameters+5ModelScoreCostPer taskWhat changes
MiMo v2.5 ProthinkingOpen×3.5-18%$0.054Memory (4-bit): 18 GB → 614 GBMiMo v2.6 FlashthinkingOpen×3.4-5%$0.062Memory (4-bit): 18 GB → 186 GB+4GPT-6 Lunaxhigh×3.3-36%$0.042Function calling only with none · Chat Completions+3Claude Haiku 5.5medium×3.3-29%$0.047Remove sampling parameters+5GPT-5.6 Lunahigh×3.2-33%$0.044ModelScoreCostPer taskWhat changes
GPT-6 Lunalow+97%-93%$0.005Function calling only with none · Chat Completions+3GPT-5.6 Lunalow+93%-85%$0.01MiMo v2.5thinkingOpen×2.6-71%$0.019Memory (4-bit): 18 GB → 186 GBClaude Haiku 5.5low×2.6-63%$0.024Remove sampling parameters+5
Mercury 2high+67%-41%$0.039Context: 262K → 128KModelScoreCostPer taskWhat changes
Qwen3.8 27BxhighOpen×3.3+1,428%$1.01max_tokens range · QwenCloud+4
Nemotron 3.5 Bolt 74B-A9BOpen×2.4——Context: 262K → 256K+1Qwen 3.5 27BthinkingOpen×2.3—$0.30 / $2.40
Gemma 4 31BthinkingOpen×2.3—$0.10 / $0.34Context: 262K → 256K+1IK2 Horizon MoVA 36B A4BthinkingOpen×2.3—$1 / $4Memory (4-bit): 18 GB → 22 GBModelScoreCostPer taskWhat changes
Qwen3.8 Max (0902)thinking×4.2+8,101%$5.41Qwen3.8 2.4T A95BthinkingOpen×4.1+3,168%$2.16Memory (4-bit): 18 GB → 1,468 GB+5Qwen3.7 Maxthinking×3.9+1,640%$1.15Qwen3.8 FlashthinkingOpen×3.7+464%$0.37Memory (4-bit): 18 GB → 108 GB+6Qwen3.7 Plusthinking×3.5+229%$0.22Qwen3.6 Max Previewthinking×3—$1.30 / $7.80Include the word JSON · QwenCloudQwen3.5 397B A17BthinkingOpen×2.8+620%$0.47Memory (4-bit): 18 GB → 242 GBQwen3 Max Thinkingthinking×2.7—$1.20 / $6Context: 262K → 256KAll Alibaba (Qwen) models →Run it yourself
estimateFits on✗ 8 GB (phone, laptop)✗ 16 GB✓ 24 GB GPU · compressed✓ 48 GB GPU · 8-bit✓ 80 GB GPU · full quality
LicenceApache-2.0free for commercial use
History
2 versionsBenchmarks
in Chat & KnowledgeAA Intelligence Index 4.3.29.8#211
AA Omniscience Index-56.1#242
AA Omniscience · Accuracy16.3#215
Humanity's Last Exam10.3#187
GPQA Diamond70.7#182
AA-LCR (long context) 1.161.3#153
IFBench (instructions)50.7#114
CritPt (physics)0.3#149
GDP.pdf · AA all-pass0.4#128
AIME 2025 · AA archived56.3#84
MMLU-Pro · official boardlab80.9
FAQ
How much does Qwen3 30B A3B 2507 (Reasoning) cost?›
There is no per-unit API price: price not verified.
How good is Qwen3 30B A3B 2507 (Reasoning), and what is it best at?›
#186 of 334 in Chat & Knowledge, #75 of 132 in Coding, #114 of 133 in Agents & Tool Use, #113 of 245 in Writing & Language, #128 of 133 in Vision & Documents.
Is Qwen3 30B A3B 2507 (Reasoning) better than Qwen3 14B?›
On Humanity's Last Exam, Qwen3 30B A3B 2507 (Reasoning) scores 10.3 vs 4.5 — higher.
What is the context window of Qwen3 30B A3B 2507 (Reasoning)?›
262K tokens.
When was Qwen3 30B A3B 2507 (Reasoning) released? Is there a newer version?›
Released 2025-07-30. It has been superseded by Qwen 3.5 35B A3B.
Is Qwen3 30B A3B 2507 (Reasoning) open source?›
Yes, the weights are open (Apache-2.0 licence).
Can I run Qwen3 30B A3B 2507 (Reasoning) locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can Qwen3 30B A3B 2507 (Reasoning) read images?›
No, it accepts text only.
What is Qwen3 30B A3B 2507 (Reasoning)?›
Qwen3 30B A3B 2507 (Reasoning) is an open-weights language model from Alibaba (Qwen), released Jul 2025, the successor to Qwen3 14B.
What are the best alternatives to Qwen3 30B A3B 2507 (Reasoning)?›
MiMo v2.5 Pro (cheaper), GPT-6 Luna (better), Qwen3.8 27B (open weights).