Open-weight model
Qwen3-30B-A3B-Instruct-2507_53-MoBE
by Zhendong Mi zmi/Qwen3-30B-A3B-Instruct-2507_53-MoBE
Qwen3-30B-A3B-Instruct-2507_53-MoBE is an open-weight model from Zhendong Mi. It has 20.3B parameters and a 262,144-token context. At 16-bit it needs about 48.6 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index.
Runs On
What it takes to serve Qwen3-30B-A3B-Instruct-2507_53-MoBE (20.3B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.
| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
|---|---|---|---|---|---|
| 16-bit | 40.5 GB | 48.6 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 20.3 GB | 24.3 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 10.1 GB | 12.2 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Oct 20, 2026.
Qwen3-30B-A3B-Instruct-2507_53-MoBE on every accelerator the SAVRN Index prices, at every precision
Model Card
The publisher has not written a card for this model.
Configuration
- Architecture
- Qwen3MoeForCausalLM
- Context length (tokens)
- 262,144
- Layers
- 48
- Hidden size
- 2,048
- Feed-forward size
- 6,144
- Attention heads
- 32
- Key/value heads
- 4
- Head dimension
- 128
- Vocabulary size
- 151,936
- Experts
- 128
- Experts active per token
- 8
- RoPE base
- 10,000,000
- Stored precision
- bfloat16
- Model type
- qwen3_moe
Identity and Version
- Repository
- zmi/Qwen3-30B-A3B-Instruct-2507_53-MoBE
- Publisher
- Zhendong Mi
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- 20.3B parameters
- Languages
- Not stated by the source
- Revision
- 57108e83775456ba35b2131d72d61760860e8b23
- First published
- 2026-10-13
- Last updated
- 2026-10-13
Files and Weights
17 files, 40.6 GB in total. The weights are 7 files totalling 40.5 GB in safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model-00001-of-00007.safetensors | Weights | 5.9 GB | c5ab2a1a1d4e |
| model-00002-of-00007.safetensors | Weights | 6.0 GB | 534e810b0279 |
| model-00003-of-00007.safetensors | Weights | 6.0 GB | cfd608f2de21 |
| model-00004-of-00007.safetensors | Weights | 6.0 GB | 2358cf74601e |
| model-00005-of-00007.safetensors | Weights | 6.0 GB | d6c651f65502 |
| model-00006-of-00007.safetensors | Weights | 6.0 GB | 85dca4e777be |
| model-00007-of-00007.safetensors | Weights | 4.7 GB | 60c58ae43a49 |
| added_tokens.json | Configuration | 707 B | — |
| config.json | Configuration | 1.0 KB | — |
| generation_config.json | Configuration | 213 B | — |
| model.safetensors.index.json | Configuration | 1.9 MB | — |
| special_tokens_map.json | Configuration | 613 B | — |
| .gitattributes | Repository | 1.6 KB | — |
| merges.txt | Tokenizer | 1.7 MB | — |
| tokenizer.json | Tokenizer | 11.4 MB | aeb13307a71a |
| tokenizer_config.json | Tokenizer | 8.2 KB | — |
| vocab.json | Tokenizer | 2.8 MB | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 40.5 GB
Released by Zhendong Mi through its official repository on Hugging Face.
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 40.5 GB |
| 16-bit | 40.5 GB |
| 8-bit | 20.3 GB |
| 4-bit | 10.1 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About Qwen3-30B-A3B-Instruct-2507_53-MoBE
How much GPU memory does Qwen3-30B-A3B-Instruct-2507_53-MoBE need?
About 48.6 GB at 16-bit and 12.2 GB at 4-bit: the weights (20.3B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run Qwen3-30B-A3B-Instruct-2507_53-MoBE on?
At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
What is Qwen3-30B-A3B-Instruct-2507_53-MoBE's context length?
262,144 tokens, from the maximum position embeddings in its published configuration.