Capability matrix
- Compute and format Q4_K_M · Intel and AMD x86 vision/parity lanes.
- Local entry point Image + prompt.
- Server Unverified for this artifact. Check for a service that accepts this artifact.
- Harness No recorded client task. Run one identified request, then a task with a checked result.
Choose the download
Publisher model card · CKE input files
Publisher task: image text to text. CKE example: Q4_K_M GGUF. Use this artifact identifier:
hf://ggml-org/Qwen3.6-27B-GGUF/Qwen3.6-27B-Q4_K_M.gguf.
Get started
Add --mmproj hf://ggml-org/Qwen3.6-27B-GGUF/mmproj-Qwen3.6-27B-Q8_0.gguf; use the separate OCR command in the v8 runbook because text-only execution does not certify vision Use the model-specific setup.
What has actually run?
Qwen3-VL is the promoted vision baseline; Qwen3.6-VL has a separate resumable private vision-evaluation lane. 1,024-token text smoke plus image-corpus checkpoints.
CPU evidence. Intel and AMD x86 vision/parity lanes; exact scope is report-specific.
Performance. The cited model result does not give a comparable prefill/decode pair here. ARM throughput for this exact artifact is unverified here. Benchmark evidence · 32K to 128K results
Can I fine-tune it?
Full-checkpoint training, LoRA and other PEFT methods: no verified CKE recipe for this exact checkpoint in the reviewed evidence. For a trainable starter, use the FP32 and adapter recipes.
Configure the OCR trial
Prepare the matching decoder GGUF, generated encoder runtime and image evaluation manifest. Point CK_QWEN36VL_DECODER_GGUF, CK_QWEN36VL_ENCODER_RUNTIME and CK_QWEN36VL_OCR_MANIFEST at those files. A language-only response cannot validate an image. For a first public image example, use Qwen3-VL with its included test card.