Qwen3 VL 235B A22B Instruct
Sep 2025
Older versionReplace withQwen3.5 397B A17BNewer version
MiniMax M3.1 FlashCheaper, as good
GPT-6 LunaBetter
GLM 5.3 FlashOpen weights
Qwen3.5 397B A17BNewer versionMiniMax M3.1 FlashCheaper, as goodGPT-6 LunaBetterGLM 5.3 FlashOpen weightsQwen3 VL 235B A22B InstructCompare inon
Benchmarks of Qwen3 VL 235B A22B Instruct32
BenchmarkByScoreRank
Overall1
Knowledge4
Science2
Coding1
Multimodal2
Writing15
Documents1
Legacy · saturated or no longer reported6
No benchmark matches
Rankingin non-thinking mode?
Retired models are left out of the current rankings.
Alternatives?
ModelScoreCostPer taskWhat changes
CheaperClaude Haiku 5.5max×4.4-46%$0.21Remove sampling parameters+5BetterGPT-6 Lunalow×2.2-46%$0.005Function calling only with none · Chat Completions+3OpenGLM 5.3 Flashmax×4.2-36%$0.28Memory (4-bit): 141 GB → 193 GB+5CheaperMiniMax M3.1 Flashthinking×4.3-29%$0.22Complete model response must be appendedCheaperQwen3.8 FlashthinkingOpen×4-38%$0.37max_tokens range · QwenCloud+5ModelScoreCostPer taskWhat changes
Claude Haiku 5.5max×4.4-46%$0.21Remove sampling parameters+5MiniMax M3.1 Flashthinking×4.3-29%$0.22Complete model response must be appendedGLM 5.3 FlashmaxOpen×4.2-36%$0.28Memory (4-bit): 141 GB → 193 GB+5Qwen3.8 FlashthinkingOpen×4-38%$0.37max_tokens range · QwenCloud+5GPT-6 Lunamax×3.8-46%$0.068Function calling only with none · Chat Completions+3ModelScoreCostPer taskWhat changes
GPT-6 Lunalow×2.2-46%$0.005Function calling only with none · Chat Completions+3GPT-5.6 Lunalow×2.1+22%$0.01
MiMo v2.5thinkingOpen×2.2-53%$0.019Memory (4-bit): 141 GB → 186 GBClaude Haiku 5.5low×3-46%$0.024Remove sampling parameters+5GPT-5 Minihigh+69%+86%$0.053ModelScoreCostPer taskWhat changes
GLM 5.3 FlashmaxOpen×4.2-36%$0.28Memory (4-bit): 141 GB → 193 GB+5
Ling 3.1 FlashOpen×4.1+22%$1.00Memory (4-bit): 141 GB → 343 GB+1Qwen3.8 FlashthinkingOpen×4-38%$0.37max_tokens range · QwenCloud+5MiMo v2.6 FlashthinkingOpen×3.8-53%$0.062Memory (4-bit): 141 GB → 186 GB+4Qwen3.8 27BxhighOpen×3.4+158%$1.01max_tokens range · QwenCloud+4ModelScoreCostPer taskWhat changes
Qwen3.8 Max (0902)thinking×4.6—$5.41Qwen3.8 2.4T A95BthinkingOpen×4—$2.16Memory (4-bit): 141 GB → 1,468 GB+6Qwen3.8 FlashthinkingOpen×4—$0.37max_tokens range · QwenCloud+5Qwen3.7 Maxthinking×3—$1.15No image inputQwen3.6 Max Previewthinking×2.9—$1.30 / $7.80No image input+1Qwen3.8 27BmediumOpen×2.8—$1.13max_tokens range · QwenCloud+4Qwen3.6 Plusthinking×2.7—$0.50 / $3Qwen3.7 Plusthinking×2.5—$0.22All 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
AA Omniscience Index-52.7
GPQA Diamond71.2
Coding
Agents & Tool Use
τ²-bench · AA35.1
Writing & Language
Vision & Documents
FAQ
How much does Qwen3 VL 235B A22B Instruct cost?›
$0.20 per 1M input tokens and $0.88 per 1M output tokens (cheapest host: DeepInfra).
Is Qwen3 VL 235B A22B Instruct free to use?›
There is no free API; the cheapest paid price is $0.20 per 1M input tokens. The weights are open, so you can run it yourself.
How can I use Qwen3 VL 235B A22B Instruct?›
Through the API of DeepInfra, Parasail, Venice.
What is the context window of Qwen3 VL 235B A22B Instruct?›
262K tokens, with up to 33K output tokens.
When was Qwen3 VL 235B A22B Instruct released? Is there a newer version?›
Released 2025-09-23. It has been superseded by Qwen3.5 397B A17B.
Is Qwen3 VL 235B A22B Instruct open source?›
Yes, the weights are open (Apache-2.0 licence).
Can I run Qwen3 VL 235B A22B Instruct locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can Qwen3 VL 235B A22B Instruct read images?›
It accepts text and images as input.
What is Qwen3 VL 235B A22B Instruct?›
Qwen3 VL 235B A22B Instruct is an open-weights language model from Alibaba (Qwen), released Sep 2025.
What are the best alternatives to Qwen3 VL 235B A22B Instruct?›
Claude Haiku 5.5 (cheaper), GPT-6 Luna (better), GLM 5.3 Flash (open weights).