JevFinetune is an open-weight model from Joebu. It has 5.2B parameters and a 262,144-token context. At 16-bit it needs about 12.4 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index.
Multimodal JEV is a visual judgment model based on Qwen3.5-4B. It evaluates questions about images and returns structured judgments and probabilities rather than generating free-form answers.
Runs On
What it takes to serve JevFinetune (5.2B 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 | 10.4 GB | 12.4 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 5.2 GB | 6.2 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 2.6 GB | 3.1 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.
JevFinetune on every accelerator the SAVRN Index prices, at every precision
Model Card
Multimodal JEV is a visual judgment model based on Qwen3.5-4B. It evaluates questions about images and returns structured judgments and probabilities rather than generating free-form answers. The checkpoint includes model weights, decision and ordinal heads, tokenizer, image processor, configuration and calibration metadata. Weights are stored in six indexed safetensors shards, totaling approximately 9.7 GiB. Inference requires the custom UnifiedJev loader, the mso runtime modules and their Python dependencies. These must be available on your Python import path; downloading the checkpoint alone is not sufficient. Standard Qwen loading does not implement the JEV judgment interface. The…
Excerpt from the card by Joebu.
Configuration
- Architecture
- Qwen3_5ForConditionalGeneration
- Context length (tokens)
- 262,144
- Layers
- 32
- Hidden size
- 2,560
- Feed-forward size
- 9,216
- Attention heads
- 16
- Key/value heads
- 4
- Head dimension
- 256
- Vocabulary size
- 248,079
- Model type
- qwen3_5
Identity and Version
- Repository
- joebumodels/JevFinetune
- Publisher
- Joebu
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- 5.2B parameters
- Languages
- en, zh
- Revision
- f6e971e250211e3d13b8e77ed96b9017aa10ebab
- First published
- 2026-10-11
- Last updated
- 2026-10-12
Files and Weights
33 files, 10.4 GB in total. The weights are 6 files totalling 10.4 GB in safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model-00001.safetensors | Weights | 2.0 GB | 74341afb98e9 |
| model-00002.safetensors | Weights | 2.0 GB | fea919ca51f6 |
| model-00003.safetensors | Weights | 2.0 GB | 8baaa02185e5 |
| model-00004.safetensors | Weights | 2.0 GB | 7aecfa6ae13b |
| model-00005.safetensors | Weights | 1.2 GB | 516bfb918ba6 |
| model-00006.safetensors | Weights | 1.3 GB | c0109b3a277f |
| config.json | Configuration | 2.9 KB | — |
| head_meta.json | Configuration | 1.6 KB | — |
| jev_config.json | Configuration | 176 B | — |
| model.safetensors.index.json | Configuration | 67.5 KB | — |
| mso/__init__.py | Configuration | — | |
| mso/action.py | Configuration | 7.7 KB | — |
| mso/action_infer.py | Configuration | 4.8 KB | — |
| mso/branch.py | Configuration | 11.4 KB | — |
| mso/fast_kernels.py | Configuration | 2.6 KB | — |
| mso/head.py | Configuration | 7.3 KB | — |
| mso/infer.py | Configuration | 23.5 KB | — |
| mso/panels.py | Configuration | 3.5 KB | — |
| mso/records.py | Configuration | 11.2 KB | — |
| mso/templates.py | Configuration | 10.5 KB | — |
| mso/v04.py | Configuration | 2.8 KB | — |
| mso/video.py | Configuration | 3.6 KB | — |
| processor_config.json | Configuration | 1.2 KB | — |
| unified_loader.py | Configuration | 23.2 KB | — |
| LICENSE-OmniJev | Documentation | 11.7 KB | — |
| LICENSE-Qwen | Documentation | 11.3 KB | — |
| LICENSE-runtime | Documentation | 11.7 KB | — |
| README.md | Documentation | 3.2 KB | — |
| chat_template.jinja | Other | 7.8 KB | — |
| requirements.txt | Other | 210 B | — |
| .gitattributes | Repository | 1.6 KB | — |
| tokenizer.json | Tokenizer | 20.0 MB | 2f838eae798e |
| tokenizer_config.json | Tokenizer | 1.3 KB | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 10.4 GB
Released by Joebu through its official repository on Hugging Face.
Built From
- Derived from Qwen/Qwen3.5-4B
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 10.4 GB |
| 16-bit | 10.4 GB |
| 8-bit | 5.2 GB |
| 4-bit | 2.6 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About JevFinetune
How much GPU memory does JevFinetune need?
About 12.4 GB at 16-bit and 3.1 GB at 4-bit: the weights (5.2B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run JevFinetune 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 JevFinetune's context length?
262,144 tokens, from the maximum position embeddings in its published configuration.