Qwen
Qwen
Multimodal model · China ·May 2026

Qwen-VLA

Multimodal Open Weights
5.2B
Parameters
1
Benchmarks

About Qwen-VLA

Qwen-VLA is an AI model developed by Qwen, in the multimodal category, released in 2026, made available as an open-weights model with 5.2B params. Alibaba Qwen's first dedicated vision-language-action foundation model for embodied AI, extending the Qwen VL stack to c.

On this page you'll find Qwen-VLA's full specifications, including 1 benchmark result. Review provider pricing and benchmark scores below, or compare Qwen-VLA head-to-head with other multimodal models.

→ More models by Qwen → All multimodal models → AI benchmarks leaderboard

Run Qwen-VLA locally — what to buy

Share this build ↗

Sized to what Qwen-VLA actually needs (~8k context at Q4), not to the biggest GPU: Good = cheapest that runs it, Better = best value, Best = most headroom. Speed is a hardware estimate; anything past ~120 tok/s is shown as instant.

Good
Intel Arc B580 12GB
Consumer GPU · ~$249
~113.8 tok/s
Q4 · fits full context · up to 8k ctx
Runs with: ExLlamaV2, KoboldCpp, LM Studio
See full estimate →
Better
NVIDIA GeForce RTX 4070 Super
Consumer GPU · ~$599
~120 tok/s
Q4 · fits full context · up to 8k ctx
Runs with: ExLlamaV2, KoboldCpp, LM Studio
See full estimate →
Best
Intel Arc A770 16GB
Consumer GPU · ~$329
~120 tok/s
Q4 · fits full context · up to 8k ctx
Runs with: ExLlamaV2, KoboldCpp, LM Studio
See full estimate →

Software support

·ExLlamaV2 ·KoboldCpp ·LM Studio ·MLX ·Ollama ·SGLang ·TensorRT-LLM ·llama.cpp ·vLLM ?KTransformers ?text-generation-webui

✓ measured · · compatible · — not supported. Informational only — speed is hardware-based.

Consumer GPU
55.9–120+ tok/s
32 options →
Apple
33.8–120+ tok/s
18 options →
Datacenter GPU
120+ tok/s
13 options →

Community

Would you run Qwen-VLA again?
No votes yet — be the first.
Did it run? — community reports by hardware

No reports yet. Own this model on some hardware? Be the first to confirm it runs.

Discussion

No comments yet. Start the discussion.

Qwen-VLA benchmark scores▶
Compare on the benchmark table →
Benchmark Category Score Variant
LIBERO agentic 97.9%

More models like Qwen-VLA

AI
NeuroChess
· 72.3K
 Qwen
Qwen3.7-Plus
Qwen
 Qwen
Qwen3.5-LiveTranslate-Flash
Qwen
 Qwen
Qwen3.7-Max
Qwen
 Qwen
Qwen3.6-Max-Preview
Qwen
 Qwen
LaSER-Qwen3-8B
Qwen · 8B

Frequently asked questions

What is Qwen-VLA? ▶

Alibaba Qwen's first dedicated vision-language-action foundation model for embodied AI, extending the Qwen VL stack to c It is tracked on GenAIList with its specifications, benchmark scores and provider pricing.

Who created Qwen-VLA? ▶

Qwen-VLA was developed by Qwen and released in 2026.

Is Qwen-VLA open source or proprietary? ▶

Qwen-VLA ships as an open-weights model: you can download and self-host the weights, though the licence may place some restrictions on use.

How much does Qwen-VLA cost? ▶

Pricing for Qwen-VLA depends on the provider. See the providers table on this page for the latest API rates.

How does Qwen-VLA perform on benchmarks? ▶

Qwen-VLA is benchmarked across 1 evaluation on GenAIList, including LIBERO (97.9%). See the full benchmark table below and compare it with other models.

Qwen-VLA Explained: One 4K Model for Robot Actions
Qwen-VLA: A Unified Foundation Model for Embodied AI
Qwen-VLA: Unifying Vision-Language-Action Modeling across Tasks and Robot Embodiments

Know a great video about Qwen-VLA?