Skip to content
ModelListedExternal
Download

Caid Technologies Parti Vision GGUF

> Parti-Vision Base is a Qwen3.5-9B fine-tune by caid-technologies that turns a hardware idea — a sentence, a brief, or a sketch — into a complete build blueprint, returning a single structured JSON object containing...

KindOtherBase modelcaid-technologies/parti-visionVersionRevision 9fef0d7Bgrade

Model source

Kind
Other
File size
595 MB–16.7 GB
Base model
caid-technologies/parti-vision
Version
vRevision 9fef0d7
Updated
Sep 26, 2026
Downloads
192
Source
Hugging Face

Source description

caid-technologies-parti-vision-GGUF

Parti-Vision Base is a Qwen3.5-9B fine-tune by caid-technologies that turns a hardware idea — a sentence, a brief, or a sketch — into a complete build blueprint, returning a single structured JSON object containing a parts list, pin-level wiring, ordered build steps, a costed sourcing table, and an appearance spec with a ready-to-use image-generation prompt. It's a standalone, adapter-free merged 16-bit model (a LoRA-adapter version is also available) trained via Unsloth on 150 synthetic schema- and semantic-gate-validated records across four input variants (prompt-only, plus brief, plus sketch, plus brief+render), accepting plain-English requests, pasted documents, or concept images/sketches as vision input, and defaulting to direct JSON output with enable_thinking=False (though a reasoning trace can be enabled). While the stock base Qwen3.5-9B scores 0% on blueprint validity, Parti-Vision achieves ~67% schema-valid in-distribution and ~61% on unseen realistic prompts with free decoding — jumping to ~83% and ~97% respectively when served with vLLM's guided_json constrained decoding — though it remains an early research preview where roughly half of outputs contain at least one design slip (wrong pin nets, cost errors, misordered steps), making it a tool for rapid ideation and drafting rather than a substitute for real engineering, CAD, or safety review.

Model Files

File NameQuant TypeFile SizeFile Link
caid-technologies-parti-vision.BF16.ggufBF1617.9 GBDownload
caid-technologies-parti-vision.F16.ggufF1617.9 GBDownload
caid-technologies-parti-vision.Q3_K_L.ggufQ3_K_L4.93 GBDownload
caid-technologies-parti-vision.Q3_K_M.ggufQ3_K_M4.62 GBDownload
caid-technologies-parti-vision.Q3_K_S.ggufQ3_K_S4.26 GBDownload
caid-technologies-parti-vision.Q4_K_M.ggufQ4_K_M5.63 GBDownload
caid-technologies-parti-vision.Q4_K_S.ggufQ4_K_S5.35 GBDownload
caid-technologies-parti-vision.Q5_K_M.ggufQ5_K_M6.47 GBDownload
caid-technologies-parti-vision.Q5_K_S.ggufQ5_K_S6.31 GBDownload
caid-technologies-parti-vision.Q8_0.ggufQ8_09.53 GBDownload
caid-technologies-parti-vision.mmproj-bf16.ggufmmproj-bf16922 MBDownload
caid-technologies-parti-vision.mmproj-f16.ggufmmproj-f16922 MBDownload
caid-technologies-parti-vision.mmproj-q8_0.ggufmmproj-q8_0624 MBDownload

llama.cpp

LLM inference in C/C++ — https://github.com/ggml-org/llama.cpp

Sources

1 source
Hugging FacePrimaryVersion 9fef0d745604aff824a4cc553b8fa599eaa1ec6e
huggingface.co/prithivMLmods/caid-technologies-parti-vision-GGUF/tree/9fef0d745604aff824a4cc553b8fa599eaa1ec6e

Verified Aug 28

Model artifacts

13 artifacts

caid-technologies-parti-vision.BF16.gguf

gguf · 16.7 GB · SHA-256 515fd947feb5…1aa9 · Hugging Face

Download

caid-technologies-parti-vision.F16.gguf

gguf · 16.7 GB · SHA-256 1e5de68ef608…2a8c · Hugging Face

Download

caid-technologies-parti-vision.mmproj-bf16.gguf

gguf · 879 MB · SHA-256 d0335d0f6c0d…70cc · Hugging Face

Download

caid-technologies-parti-vision.mmproj-f16.gguf

gguf · 879 MB · SHA-256 d0335d0f6c0d…70cc · Hugging Face

Download

caid-technologies-parti-vision.mmproj-q8_0.gguf

gguf · 595 MB · SHA-256 7da2e57a01bd…b647 · Hugging Face

Download

caid-technologies-parti-vision.Q3_K_L.gguf

gguf · 4.59 GB · SHA-256 c2fd3a992d8c…d577 · Hugging Face

Download

caid-technologies-parti-vision.Q3_K_M.gguf

gguf · 4.31 GB · SHA-256 0be686eaebbc…45d8 · Hugging Face

Download

caid-technologies-parti-vision.Q3_K_S.gguf

gguf · 3.97 GB · SHA-256 c917e1f86398…e1af · Hugging Face

Download

caid-technologies-parti-vision.Q4_K_M.gguf

gguf · 5.24 GB · SHA-256 70a5c56ab3d8…e604 · Hugging Face

Download

caid-technologies-parti-vision.Q4_K_S.gguf

gguf · 4.98 GB · SHA-256 f3abcd1a0128…8720 · Hugging Face

Download

caid-technologies-parti-vision.Q5_K_M.gguf

gguf · 6.02 GB · SHA-256 7a8d7cec29f0…2d41 · Hugging Face

Download

caid-technologies-parti-vision.Q5_K_S.gguf

gguf · 5.87 GB · SHA-256 3516dd529de5…2f57 · Hugging Face

Download

caid-technologies-parti-vision.Q8_0.gguf

gguf · 8.87 GB · SHA-256 9c2c7dc28a66…6b59 · Hugging Face

Download

Source excerpts

2 excerpts

--- license: apache-2.0 base_model: - caid-technologies/parti-vision library_name: transformers tags: - text-generation-inference - llama-cpp - caid - blueprint - hardware - electronics - maker - structured-generation - json - qwen3.5 - vision language: - en pipeline_tag: image-text-to-text --- # **caid-technologies-parti-vision-GGUF** > **[Parti-Vision Base](https://huggingface.co/caid-technologies/parti-vision)** is a Qwen3.5-9B fine-tune by caid-technologies that turns a hardware idea — a sentence, a brief, or a sketch — into a complete build blueprint, returning a single structured JSON object containing a parts list, pin-level wiring, ordered build steps, a costed sourcing table, and an appearance spec with a ready-to-use image-generation prompt. It's a standalone, adapter-free merged 16-bit model (a LoRA-adapter version is also available) trained via Unsloth on 150 synthetic schema- and semantic-gate-validated records across four input variants (prompt-only, plus brief, plus sketch, plus brief+render), accepting plain-English requests, pasted documents, or concept images/sketches as vision input, and defaulting to direct JSON output with `enable_thinking=False` (though a reasoning trace can be enabled). While the stock base Qwen3.5-9B scores 0% on blueprint validity, Parti-Vision achieves ~67% schema-valid in-distribution and ~61% on unseen realistic prompts with free decoding — jumping to ~83% and ~97% respectively when served with vLLM's `guided_json` constrained decoding — though it remains an early research preview where roughly half of outputs contain at least one design slip (wrong pin nets, cost errors, misordered steps), making it a tool for rapid ideation and drafting rather than a substitute for real engineering, CAD, or safety review. ## Model Files File Name | Quant Type | File Size | File Link | |-----------|------------|-----------|-----------| | caid-technologies-parti-vision.BF16.gguf | BF16 | 17.9 GB | [Download](https://huggingface.co/prithivMLmods/caid-technologies-parti-vision-GGUF/blob/main/caid-technologies-parti-vision.BF16.gguf) | | caid-technologies-parti-vision.F16.gguf | F16 | 17.9 GB | [Download](https://huggingface.co/prithivMLmods/caid-technologies-parti-vision-GGUF/blob/main/caid-technologies-parti-vision.F16.gguf) | | caid-technologies-parti-vision.Q3_K_L.gguf | Q3_K_L | 4.93 GB | [Download](https://huggingfa...

Source context: 192 downloads · 1 likes · Pipeline image-text-to-text · Library transformers · Repo prithivMLmods/caid-technologies-parti-vision-GGUF