Qwen 3 VL 235B A22B Thinking
Sep 2025
Qwen 3 VL 235B A22B ThinkingCompare inon
Benchmarks of Qwen 3 VL 235B A22B Thinking14
BenchmarkByScoreRank
Multimodal1
Writing13
No benchmark matches
Rankingconfiguration unstated?
Alternatives?
ModelScoreCostPer taskWhat changes
CheaperGemini 3 Flash Previewunstatedwins 66%-7%$0.50 / $3
BetterGPT-6 Lunamaxwins 60%-83%$0.068Function calling only with none · Chat Completions+3
OpenGLM 5.3 Flashunstatedwins 63%-80%$0.15 / $0.50Memory (4-bit): 141 GB → 193 GB+5
CheaperMiMo v2.6 ProunstatedOpenwins 65%-55%$0.43 / $0.87Memory (4-bit): 141 GB → 615 GB+4CheaperGemini 3.5 Flash Liteunstatedwins 64%-30%$0.30 / $2.50ModelScoreCostPer taskWhat changes
Gemini 3 Flash Previewunstatedwins 66%-7%$0.50 / $3MiMo v2.6 ProunstatedOpenwins 65%-55%$0.43 / $0.87Memory (4-bit): 141 GB → 615 GB+4Gemini 3.5 Flash Liteunstatedwins 64%-30%$0.30 / $2.50GLM 5.3 FlashunstatedOpenwins 63%-80%$0.15 / $0.50Memory (4-bit): 141 GB → 193 GB+5MiMo v2.5 ProunstatedOpenwins 63%-55%$0.43 / $0.87Memory (4-bit): 141 GB → 614 GB+1ModelScoreCostPer taskWhat changes
GPT-6 Lunamaxwins 60%-83%$0.068Function calling only with none · Chat Completions+3GPT-5.6 Lunaxhighwins 60%-63%$0.085GPT-5.6 Terraxhighwins 63%+272%$0.63Prompts with >272K input tokensGPT-6.1 Solmaxwins 67%+231%$0.72Remove unsupported parameters+5
Muse Spark 1.2xhighwins 65%+65%$0.97Only auto tool_choice is supported+1ModelScoreCostPer taskWhat changes
GLM 5.3 FlashunstatedOpenwins 63%-80%$0.15 / $0.50Memory (4-bit): 141 GB → 193 GB+5
Hy3unstatedOpenwins 62%-81%$0.13 / $0.53Context: 262K → 256K+2Gemma 4 31BunstatedOpenwins 61%-87%$0.10 / $0.34Context: 262K → 256KQwen3.5 397B A17BunstatedOpenwins 60%+12%$0.60 / $3.60Memory (4-bit): 141 GB → 242 GB
MiniMax M3unstatedOpenwins 60%-57%$0.30 / $1.20Memory (4-bit): 141 GB → 256 GB+2ModelScoreCostPer taskWhat changes
Qwen 3.7 Max Previewwins 64%—$2.50 / $7.50Context: not verified+1Qwen3.6 Max Previewunstatedwins 64%—$1.30 / $7.80No image input+1Qwen3.7 Plusunstatedwins 63%—$0.32 / $1.28Qwen3.6 Plusunstatedwins 60%—$0.50 / $3Qwen3.5 397B A17BunstatedOpenwins 60%—$0.60 / $3.60Memory (4-bit): 141 GB → 242 GBQwen3 Max (Preview)unstatedwins 58%—$0.87 / $3.48No image inputQwen 3Openwins 58%——Context: not verified+2Qwen3 235B A22B Thinking 2507unstatedOpenwins 55%—$0.23 / $2.30No image input+1All Alibaba (Qwen) models →Run it yourself
estimateFits on✗ 8 GB (phone, laptop)✗ 16 GB✗ 24 GB GPU✗ 48 GB GPU✗ 80 GB GPU
LicenceApache-2.0free for commercial use
Benchmarks
all categoriesChat & Knowledge
Arena Text · overall1,395#103
Arena Text · Chinese1,441.6#88
Arena Text · English1,420.1#83
Arena Text · German1,417.1#62
Arena Text · Korean1,358#73
Arena Text · non-English1,377#96
Arena Text · Polish1,365.6#86
Arena Text · Russian1,370.3#99
Arena Text · Spanish1,402.6#85
Arena Text · overall · full cohort1,400.7#89
Writing & Language
Arena Text · creative writing1,338#106
Vision & Documents
Arena Vision · overall1,207#63
FAQ
How much does Qwen 3 VL 235B A22B Thinking cost?›
$0.45 per 1M input tokens and $3.49 per 1M output tokens (cheapest host: DeepInfra).
Is Qwen 3 VL 235B A22B Thinking free to use?›
There is no free API; the cheapest paid price is $0.45 per 1M input tokens. The weights are open, so you can run it yourself.
How good is Qwen 3 VL 235B A22B Thinking, and what is it best at?›
#103 of 196 in Chat & Knowledge, #106 of 195 in Writing & Language, #63 of 88 in Vision & Documents.
Is Qwen 3 VL 235B A22B Thinking better than Qwen 3.5 Flash?›
On Arena Text · creative writing, Qwen 3 VL 235B A22B Thinking scores 1,338 vs 1,338 — about the same.
How can I use Qwen 3 VL 235B A22B Thinking?›
Through the API of DeepInfra.
What is the context window of Qwen 3 VL 235B A22B Thinking?›
262K tokens.
When was Qwen 3 VL 235B A22B Thinking released? Is there a newer version?›
Released 2025-09-22. It is the current version in its line.
Is Qwen 3 VL 235B A22B Thinking open source?›
Yes, the weights are open (Apache-2.0 licence).
Can I run Qwen 3 VL 235B A22B Thinking locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can Qwen 3 VL 235B A22B Thinking read images?›
It accepts text and images as input.
What is Qwen 3 VL 235B A22B Thinking?›
Qwen 3 VL 235B A22B Thinking is an open-weights language model from Alibaba (Qwen), released Sep 2025.
What are the best alternatives to Qwen 3 VL 235B A22B Thinking?›
Gemini 3 Flash Preview (cheaper), GPT-6 Luna (better), GLM 5.3 Flash (open weights).