rariruluis/gemma-4-12B-it-qat-OptiQ-4bit-MTP
image-text-to-text model by rariruluis
From the publisher
Model card & documentation
Source previewRead the publisher’s intended use, setup instructions, evaluations and limitations. The original model card is the source of truth.
This checkpoint combines the existing Gemma 4 12B QAT OptiQ MLX target and its matching QAT assistant into one downloadable model directory. The assistant is embedded as languagemodel.mtp. tensors in model-mtp.safetensors; there is no separate drafter download. This is an assembly and runtime-validation release, not a new fine-tune or a new quantization. The original target, tokenizer, chat template, and vision-sidecar tensor values are unchanged. All 48 assistant tensors retain their original BF16 values. Source revisions and modifications are recorded in provenance/sources.json; the original model cards are preserved alongside it. Use oMLX with its Gemma 4 Unified Lightning MTP support. The integration format relies on oMLX's pre-load runtime patches: stock Transformers or stock mlx-lm is not claimed to load this integrated checkpoint. The reference runtime is oMLX 0.7.0rc1, MLX 0.32.2, and the exact mlx-lm / mlx-vlm commits in the provenance manifest. The matching assistant configuration is embedded at textconfig.mtpassistantconfig, with textconfig.mtpnumhiddenlayers:…
Read the full model card ↗ · Preview checked 2026-10-01T08:45:25.293Z
Inside the original model card — Document outline
- Gemma 4 12B QAT OptiQ 4-bit with an integrated MTP assistant
- Runtime requirements
- Download and serve
- What was changed
- Benchmark scope and limitations
- Integrated Lightning MTP results
- SSD capacity matters for long-context prefix reuse
- Reproduce generation measurements
- License and attribution
Headings are captured from the source. Links open the publisher’s document, not a locally hosted copy.
Documentation belongs to its respective authors. Reported project/model license: apache-2.0. A listing is not a grant of reuse or training rights. Confirm the document’s own terms at the source.
Model overview
rariruluis/gemma-4-12B-it-qat-OptiQ-4bit-MTP is a image-text-to-text model repository published by rariruluis on Hugging Face. The source reports the mlx library.
This page summarizes Hub metadata. For intended use, training data, evaluation results and limitations, consult the original model card.
Model facts
- Task
- image-text-to-text
- Library
- mlx
- Recent downloads (30 days)
- 0
- Cumulative likes
- 0
- Hugging Face trending score
- 0
- Reported safetensors parameters
- 11,907,350,272
- Architecture
- Gemma4UnifiedForConditionalGeneration
- License
- apache-2.0
- Access
- Not gated by Hugging Face
- Created
- 2026-10-01T06:28:48.000Z
- Last modified
- 2026-10-01T06:29:20.000Z
Compatibility and lineage
Language coverage was not reported.
Reported base models: mlx-community/gemma-4-12B-it-qat-OptiQ-4bit, mlx-community/gemma-4-12B-it-qat-assistant-bf16
Parameter count is not a RAM/VRAM requirement. Check precision, quantization, context length and runtime compatibility in the model card; no hardware or API-cost claim is inferred here.
Use and evaluate
Check license terms and access requirements first. Review the model card’s documented loading instructions, evaluations and safety limitations. Benchmark scores are not imported by this directory, and popularity is not an accuracy ranking.
No model weights are downloaded or executed by AltAPIs. Never enable remote model code without reviewing it.
Source and freshness
Source: Hugging Face Hub. Metadata observed 2026-10-01T06:38:27.225Z. Daily imports are snapshots, not real-time monitoring.
Popularity and source listings do not establish security, suitability, licensing rights or benchmark performance.