Qwen3-Omni-30B-A3B-Captioner
- Open weights
Prices
0 offerings · prices as listedNo priced offering is listed for this model yet.
Details
- Family
- qwen-omni-captioner
Not stated on Alibaba Qwen’s pages: max input, input modalities, output modalities, reasoning, reasoning efforts, reasoning can be off, max thinking budget, tool use, structured output, built-in tools, knowledge cutoff, licence, aliases.
Run locally
Weights: Qwen/Qwen3-Omni-30B-A3B-Captioner- Parameters 31.7B
- Active 4.5B (estimated) · 128 experts, 8 per token
- Native precision BF16
- Layers 48
- KV cache per token 96 KB
- Config positions 64K (config.json max_position_embeddings)
- Needs
- ≈ 18.2 GB (weights 16.6 + KV cache 0.8 + overhead 0.8)
- Usable
- 21.6 GB (90% of its memory)
- Generation speed
- 150–225 tok/s (reads the active 4.5B per token)
- Longest context
- 44K (longer doesn’t fit)
- First token
- about 0.15–0.22 s for a 2K prompt
| Precision | Weights | Needs at 8K | On your hardware | Longest context |
|---|---|---|---|---|
| BF16 (native)maker’s file | 59.1 GB | ≈ 62.8 | ○Needs system memory 11–16 tok/s | — |
| Q8_0 | 31.4 GB | ≈ 33.7 | ○Needs system memory 21–31 tok/s | — |
| FP8 | 29.5 GB | ≈ 31.8 | ○Needs system memory 24–36 tok/s | — |
| Q6_K | 24.3 GB | ≈ 26.2 | ○Needs system memory 43–65 tok/s | — |
| Q5_K_M | 21.0 GB | ≈ 22.8 | ○Needs system memory 95–143 tok/s | — |
| Q5_0 | 20.3 GB | ≈ 22.1 | ○Needs system memory 113–170 tok/s | — |
| Q4_K_M | 17.9 GB | ≈ 19.6 | ●Runs well 141–212 tok/s | 29K |
| Q4_0 | 16.6 GB | ≈ 18.2 | ●Runs well 150–225 tok/s | 44K |
| IQ4_XS | 15.7 GB | ≈ 17.2 | ●Runs well 157–235 tok/s | 54K |
| Q3_K_Mquality drops | 14.4 GB | ≈ 15.9 | ●Runs well 167–250 tok/s | 64K |
| IQ3_XXSquality drops | 11.3 GB | ≈ 12.6 | ●Runs well 198–297 tok/s | 64K |
| Q2_Kquality drops | 11.1 GB | ≈ 12.4 | ●Runs well 201–301 tok/s | 64K |
| IQ2_XXSquality drops | 7.6 GB | ≈ 8.9 | ●Runs well 255–383 tok/s | 64K |
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
The maker publishes no files at this precision. Choose its native precision or one of its own builds, or quantise it yourself with llama.cpp.
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 →- New language model
Sources
- 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 Nov 4, 2026
- Qwen/Qwen3-Omni-30B-A3B-Captioner config.json (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3-Omni-30B-A3B-Captioner model card (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3-Omni-30B-A3B-Captioner files (opens in a new tab) huggingface.co · checked Nov 2, 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