Emomni-v1-bnb-4bit is an open-weight model from Jotakak, released under Apache License 2.0. It has 8.3B parameters. At 16-bit it needs about 20 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index. It draws 9 downloads a month.
Complete checkpoint with global LoRA already merged, followed by bitsandbytes NF4 quantization with double quantization and BF16 compute. No separate adapter is required.
Runs On
What it takes to serve Emomni-v1-bnb-4bit (8.3B 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 | 16.7 GB | 20.0 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 8.3 GB | 10.0 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 4.2 GB | 5.0 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 23, 2026.
Emomni-v1-bnb-4bit on every accelerator the SAVRN Index prices, at every precision
Model Card
By Jotakak, published under apache-2.0, revision 7d793dfb37b7.
Complete checkpoint with global LoRA already merged, followed by bitsandbytes NF4 quantization with double quantization and BF16 compute. No separate adapter is required. lmhead is excluded from quantization; existing quantization coverage of the other components is preserved. Corresponding merged BF16 checkpoint. The existing quantized weights are retained: recomputing NF4 from the matching merged BF16 checkpoint verified all 2,792 serialized tensors as identical. Use the Emomni-v1 code: The custom model is loaded by src.modelingemomni.EmomniModel, rather than the generic AutoModelForCausalLM loader. Local text/audio loading and generation were checked. These checks are not a new quality…
Read Jotakak's full model card
Emomni-v1 Base — NF4
Complete checkpoint with global LoRA already merged, followed by bitsandbytes
NF4 quantization with double quantization and BF16 compute. No separate adapter
is required. lm_head is excluded from quantization; existing quantization
coverage of the other components is preserved.
Corresponding merged BF16 checkpoint.
The existing quantized weights are retained: recomputing NF4 from the matching merged BF16 checkpoint verified all 2,792 serialized tensors as identical.
Use the Emomni-v1 code:
bash scripts/run_serve.sh --model Jotakak/Emomni-v1-bnb-4bit
The custom model is loaded by src.modeling_emomni.EmomniModel, rather than
the generic AutoModelForCausalLM loader. Local text/audio loading and generation
were checked. These checks are not a new quality benchmark.
Configuration
- Architecture
- EmomniModel
- Model type
- emomni-v1
- Quantization
- bitsandbytes
Identity and Version
- Repository
- Jotakak/Emomni-v1-bnb-4bit
- Publisher
- Jotakak
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- 8.3B parameters
- Languages
- Not stated by the source
- Revision
- 7d793dfb37b795261affaeb8893f0eee8f0b6a8d
- First published
- 2026-03-04
- Last updated
- 2026-10-11
Files and Weights
16 files, 6.1 GB in total. The weights are 2 files totalling 6.0 GB in safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model-00001-of-00002.safetensors | Weights | 4.8 GB | e0d33a22adf9 |
| model-00002-of-00002.safetensors | Weights | 1.2 GB | 59d0a6b84a4f |
| added_tokens.json | Configuration | 605 B | — |
| config.json | Configuration | 6.6 KB | — |
| generation_config.json | Configuration | 270 B | — |
| model.safetensors.index.json | Configuration | 279.5 KB | — |
| preprocessor_config.json | Configuration | 357 B | — |
| quantization_config.json | Configuration | 326 B | — |
| special_tokens_map.json | Configuration | 613 B | — |
| README.md | Documentation | 1.1 KB | — |
| chat_template.jinja | Other | 2.5 KB | — |
| .gitattributes | Repository | 1.6 KB | — |
| merges.txt | Tokenizer | 1.7 MB | — |
| tokenizer.json | Tokenizer | 11.4 MB | 9c5ae00e602b |
| tokenizer_config.json | Tokenizer | 4.7 KB | — |
| vocab.json | Tokenizer | 2.8 MB | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
- Download size
- 6.0 GB
Released by Jotakak through its official repository on Hugging Face. Read the license.
Built From
- Derived from Jotakak/Emomni-v1
- Quantized from Jotakak/Emomni-v1
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 6.0 GB |
| 16-bit | 16.7 GB |
| 8-bit | 8.3 GB |
| 4-bit | 4.2 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About Emomni-v1-bnb-4bit
How much GPU memory does Emomni-v1-bnb-4bit need?
About 20 GB at 16-bit and 5 GB at 4-bit: the weights (8.3B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run Emomni-v1-bnb-4bit 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 Emomni-v1-bnb-4bit commercially?
Yes. Emomni-v1-bnb-4bit 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.