Qwen3-VL-30B-A3B-Instruct
- In Input modalities: TextImageVideo Out Output modalities: Text
- Tool use
- Structured output
- Open weights
Prices
3 offerings · prices as listedqwen3-vl-30b-a3b-instruct| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard no longer listed since Oct 11, 2026 | $0.20 | $0.80 |
Prices from Alibaba Cloud Model Studio pricing (opens in a new tab) (checked Oct 11, 2026)
qwen3-vl-30b-a3b-instruct| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard no longer listed since Oct 11, 2026 | $0.108 | $0.431 |
Prices from Alibaba Cloud Model Studio pricing (opens in a new tab) (checked Oct 11, 2026)
qwen3-vl-30b-a3b-instruct| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard no longer listed since Oct 11, 2026 | ¥0.75 | ¥3.00 |
Prices from 阿里云百炼：模型价格 (opens in a new tab) (checked Oct 6, 2026)
Standard: the price a request with no conditions pays (the headline price elsewhere on the site).
Details
- Reasoning
- —
- Licence
- Apache-2.0
- Family
- qwen-vl
- Model ID
Not stated on Alibaba Qwen’s pages: max input, reasoning efforts, max thinking budget, built-in tools, knowledge cutoff, aliases.
Run locally
Weights: Qwen/Qwen3-VL-30B-A3B-Instruct- Parameters 31.1B
- Active 3.9B (estimated) · 128 experts, 8 per token
- Native precision BF16
- Layers 48
- KV cache per token 96 KB
- Config positions 256K (config.json max_position_embeddings)
- API context 128K (maker’s API)
- Needs
- ≈ 18.9 GB (weights 17.3 (maker’s file) + KV cache 0.8 + overhead 0.9)
- Usable
- 21.6 GB (90% of its memory)
- Generation speed
- 161–242 tok/s (reads the active 3.9B per token)
- Longest context
- 37K (longer doesn’t fit)
- First token
- about 0.13–0.19 s for a 2K prompt
| Precision | Weights | Needs at 8K | On your hardware | Longest context |
|---|---|---|---|---|
| BF16 (native)maker’s file | 57.9 GB | ≈ 61.5 | ○Needs system memory 11–17 tok/s | — |
| Q8_0maker’s file | 30.3 GB | ≈ 32.5 | ○Needs system memory 26–38 tok/s | — |
| FP8maker’s file | 30.0 GB | ≈ 32.3 | ○Needs system memory 26–39 tok/s | — |
| Q6_K | 23.8 GB | ≈ 25.7 | ○Needs system memory 54–81 tok/s | — |
| Q5_K_M | 20.6 GB | ≈ 22.4 | ○Needs system memory 128–191 tok/s | — |
| Q5_0 | 19.9 GB | ≈ 21.6 | ○Needs system memory 155–233 tok/s | — |
| Q4_K_Mmaker’s file | 17.3 GB | ≈ 18.9 | ●Runs well 161–242 tok/s | 37K |
| Q4_0 | 16.3 GB | ≈ 17.8 | ●Runs well 168–252 tok/s | 48K |
| IQ4_XS | 15.4 GB | ≈ 16.9 | ●Runs well 175–263 tok/s | 58K |
| Q3_K_Mquality drops | 14.1 GB | ≈ 15.6 | ●Runs well 186–279 tok/s | 72K |
| IQ3_XXSquality drops | 11.1 GB | ≈ 12.4 | ●Runs well 220–330 tok/s | 106K |
| Q2_Kquality drops | 10.9 GB | ≈ 12.1 | ●Runs well 223–334 tok/s | 109K |
| IQ2_XXSquality drops | 7.5 GB | ≈ 8.7 | ●Runs well 279–418 tok/s | 128K |
Click a row to choose a precision. “Maker’s file” sizes are the maker’s own files; others are parameters × bits per weight.
Compared with other hardware
Each at the highest precision it runs. Click one for details.
Smallest that runs it (Q4_K_M, 8K): GPU GeForce RTX 4090; unified memory Apple M6 (24GB or 32GB) · 32 GB.
Run commands
llama-server -hf Qwen/Qwen3-VL-30B-A3B-Instruct-GGUF:Q4_K_M -c 8192
ollama run hf.co/Qwen/Qwen3-VL-30B-A3B-Instruct-GGUF:Q4_K_M
Memory to fine-tune
Batch 1, 2,048-token sequences, gradient checkpointing; LoRA and QLoRA train rank-16 adapters on the attention projections.
Estimates; facts from the maker’s Hugging Face repositories, as of 2026-11-06. How it’s estimated
Cited scores
From third-party leaderboards, not measured by usThis model isn’t on any leaderboard we cite yet.
Changes
All changes →No changes recorded yet.
Sources
- 阿里云百炼：模型价格in Chinese (opens in a new tab) help.aliyun.com · checked Oct 6, 2026
- qwen3-vl-30b-a3b-instruct 模型信息-大模型服务平台百炼(Model Studio)-阿里云帮助中心in Chinese (opens in a new tab) help.aliyun.com · checked Oct 6, 2026
- 模型上下架与更新-大模型服务平台百炼-阿里云_大模型服务平台百炼(Model Studio)-大模型服务平台百炼(Model Studio)-阿里云帮助中心in Chinese (opens in a new tab) help.aliyun.com · checked Oct 30, 2026
- Alibaba Cloud Model Studio pricing (opens in a new tab) alibabacloud.com · checked Oct 11, 2026
- Qwen/Qwen3-VL-30B-A3B-Instruct model card (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3-VL-30B-A3B-Instruct config.json (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3-VL-30B-A3B-Instruct files (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3-VL-30B-A3B-Instruct-FP8 files (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3-VL-30B-A3B-Instruct-GGUF files (opens in a new tab) huggingface.co · checked Nov 1, 2026
- models.dev: alibaba/qwen3-vl-30b-a3b-instruct (community data, MIT) (opens in a new tab) github.com · checked Oct 29, 2026
Alibaba Qwen’s official policies: data use, privacy, terms →
Some values on this page are still waiting to be checked against official pages: a filled dot is verified, a hollow one is not yet (with or without a source on file). How we verify data
Alibaba Qwen · China · qwen.ai
Related models
Snapshots, the same family and other models from this provider