Open-weight model
Qwen3.6-35B-A3B-oQ4e-fp16-mtp
by Harish Kothandaraman hsk0201/Qwen3.6-35B-A3B-oQ4e-fp16-mtp
Qwen3.6-35B-A3B-oQ4e-fp16-mtp is an open-weight model from Harish Kothandaraman, released under Apache License 2.0. It has 36B parameters and a 262,144-token context. At 16-bit it needs about 86.3 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index.
Text and vision retained Quantised using oMLX v0.5.0.rc1 OQ Enhanced quantization (oQe) iMatrix MTP Heads retained FP16 is fastest on M1/M2, but this can work on all MLX inferencing systems This model is using the LATEST FROGGERIC chat template upgrade (Fixed…
Runs On
What it takes to serve Qwen3.6-35B-A3B-oQ4e-fp16-mtp (36B 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 | 71.9 GB | 86.3 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x MI325X $2.00 · 1x MI355X $2.59 |
| 8-bit | 36.0 GB | 43.1 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 18.0 GB | 21.6 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 18, 2026.
Qwen3.6-35B-A3B-oQ4e-fp16-mtp on every accelerator the SAVRN Index prices, at every precision
Model Card
By Harish Kothandaraman, published under apache-2.0, revision 3ddde10d21fe.
Text and vision retained Quantised using oMLX v0.5.0.rc1 OQ Enhanced quantization (oQe) iMatrix MTP Heads retained FP16 is fastest on M1/M2, but this can work on all MLX inferencing systems This model is using the LATEST FROGGERIC chat template upgrade (Fixed jinja chat templates for Qwen 3.5 & 3.6 (v21))
Read Harish Kothandaraman's full model card
Text and vision retained
Quantised using oMLX v0.5.0.rc1 OQ Enhanced quantization (oQe) iMatrix
MTP Heads retained
FP16 is fastest on M1/M2 , but this can work on all MLX inferencing systems
This model is using the LATEST FROGGERIC chat template upgrade (Fixed jinja chat templates for Qwen 3.5 & 3.6 (v21))
Configuration
- Architecture
- Qwen3_5MoeForConditionalGeneration
- Context length (tokens)
- 262,144
- Layers
- 40
- Hidden size
- 2,048
- Attention heads
- 16
- Key/value heads
- 2
- Head dimension
- 256
- Vocabulary size
- 248,320
- Experts
- 256
- Experts active per token
- 8
- Model type
- qwen3_5_moe
Identity and Version
- Repository
- hsk0201/Qwen3.6-35B-A3B-oQ4e-fp16-mtp
- Publisher
- Harish Kothandaraman
- Task
- Not stated by the source
- Modality
- Other
- Library
- mlx
- Parameters
- 36B parameters
- Languages
- mlx
- Revision
- 3ddde10d21fea97546aa2a0d59306bbaaab7f67a
- First published
- 2026-09-30
- Last updated
- 2026-09-30
Files and Weights
18 files, 22.5 GB in total. The weights are 5 files totalling 22.5 GB in safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model-00001-of-00005.safetensors | Weights | 5.1 GB | 005e20903683 |
| model-00002-of-00005.safetensors | Weights | 5.1 GB | 90ad6c9f93e7 |
| model-00003-of-00005.safetensors | Weights | 5.1 GB | 39ddbb5743d5 |
| model-00004-of-00005.safetensors | Weights | 5.1 GB | 321631a66b04 |
| model-00005-of-00005.safetensors | Weights | 2.0 GB | 136d44fc02f0 |
| config.json | Configuration | 87.7 KB | — |
| generation_config.json | Configuration | 202 B | — |
| model.safetensors.index.json | Configuration | 204.3 KB | — |
| oq_imatrix_report.json | Configuration | 32.2 KB | — |
| preprocessor_config.json | Configuration | 390 B | — |
| README.md | Documentation | 412 B | — |
| chat_template.jinja | Other | 7.8 KB | — |
| chat_template.jinja.bak | Other | 29.1 KB | — |
| .gitattributes | Repository | 1.6 KB | — |
| merges.txt | Tokenizer | 3.4 MB | — |
| tokenizer.json | Tokenizer | 12.8 MB | 5f9e4d4901a9 |
| tokenizer_config.json | Tokenizer | 16.7 KB | — |
| vocab.json | Tokenizer | 6.7 MB | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
- Download size
- 22.5 GB
Released by Harish Kothandaraman through its official repository on Hugging Face. Read the license.
Built From
- Derived from Qwen/Qwen3.6-35B-A3B
- Quantized from Qwen/Qwen3.6-35B-A3B
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 22.5 GB |
| 16-bit | 71.9 GB |
| 8-bit | 36.0 GB |
| 4-bit | 18.0 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About Qwen3.6-35B-A3B-oQ4e-fp16-mtp
How much GPU memory does Qwen3.6-35B-A3B-oQ4e-fp16-mtp need?
About 86.3 GB at 16-bit and 21.6 GB at 4-bit: the weights (36B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run Qwen3.6-35B-A3B-oQ4e-fp16-mtp 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.
Can I use Qwen3.6-35B-A3B-oQ4e-fp16-mtp commercially?
Yes. Qwen3.6-35B-A3B-oQ4e-fp16-mtp is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.
What is Qwen3.6-35B-A3B-oQ4e-fp16-mtp's context length?
262,144 tokens, from the maximum position embeddings in its published configuration.