Image-Text-to-Text
MLX
Safetensors
English
mlx-vlm
qwen
qwen3.8
distillation
multimodal
4-bit precision
8-bit precision
bf16
Instructions to use PocketAiHub/Qwen3.8-9B-MLX with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use PocketAiHub/Qwen3.8-9B-MLX with MLX:
# Make sure mlx-vlm is installed # pip install --upgrade mlx-vlm from mlx_vlm import load, generate from mlx_vlm.prompt_utils import apply_chat_template from mlx_vlm.utils import load_config # Load the model model, processor = load("PocketAiHub/Qwen3.8-9B-MLX") config = load_config("PocketAiHub/Qwen3.8-9B-MLX") # Prepare input image = ["http://images.cocodataset.org/val2017/000000039769.jpg"] prompt = "Describe this image." # Apply chat template formatted_prompt = apply_chat_template( processor, config, prompt, num_images=1 ) # Generate output output = generate(model, processor, formatted_prompt, image) print(output) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
| { | |
| "schema_version": 1, | |
| "repository": "PocketAiHub/Qwen3.8-9B-MLX", | |
| "lineage": "regular", | |
| "source_repository": "empero-ai/Qwen3.8-9B", | |
| "source_revision": "0934f3d2327ff2df2197495278c4c46ae5a56bd9", | |
| "base_model": "Qwen/Qwen3.5-9B", | |
| "runtime": { | |
| "mlx": "0.32.0", | |
| "mlx_vlm": "0.6.8" | |
| }, | |
| "format": { | |
| "precision": "4-bit", | |
| "language_quantization": { | |
| "mode": "affine", | |
| "bits": 4, | |
| "group_size": 64 | |
| }, | |
| "vision_tower": "BF16", | |
| "native_mtp_included": false, | |
| "shard_count": 2, | |
| "source_manifest": "workspace/exports/mlx/pocketai-qwen3.8-9b-mlx-4bit/artifact-manifest.json", | |
| "source_manifest_sha256": "02a89f998c02e0b0d011eed3cd69427b4041720e5554819f0676998ebb2973e1" | |
| }, | |
| "evidence": { | |
| "smoke": { | |
| "report": "workspace/runs/qwen3.8-9b-regular-benchmarks/4bit/runtime-smoke.json", | |
| "report_sha256": "91ac69d715bee134187e498842e314f5b28c381d069e8a16f2d16652f5e50d49", | |
| "result": { | |
| "passed": true, | |
| "text_passed": true, | |
| "vision_passed": true, | |
| "peak_mlx_memory_gb": 7.022197602 | |
| } | |
| }, | |
| "performance": { | |
| "report": "workspace/runs/qwen3.8-9b-regular-benchmarks/4bit-4k/report.json", | |
| "report_sha256": "dd82d29a492b1a1ce3f66540d711e9bb117fcd80b6391ef55de646418b02f638", | |
| "result": { | |
| "passed": true, | |
| "formatted_tokens": 4096, | |
| "expected": "COBALT-7319", | |
| "prefill_tokens_per_second": 3197.327651141933, | |
| "decode_tokens_per_second": 100.27767318668539, | |
| "peak_mlx_bytes": 6969058358, | |
| "kv_bits": null | |
| } | |
| }, | |
| "feature": { | |
| "report": "workspace/runs/qwen3.8-9b-release/regular-4bit-feature-64k.json", | |
| "report_sha256": "0f0c9d871812ad179e29b7c0bc4a8ea1c09bf980f09b3bb5eea540fa30fac07d", | |
| "result": { | |
| "quality": { | |
| "passed": true, | |
| "pass_count": 12, | |
| "case_count": 12 | |
| }, | |
| "tool_calling": { | |
| "passed": true, | |
| "pass_count": 8, | |
| "case_count": 8 | |
| }, | |
| "temporal_video": { | |
| "passed": true, | |
| "expected": "red->blue", | |
| "actual": "red->blue" | |
| }, | |
| "long_context": { | |
| "passed": true, | |
| "formatted_tokens": 65545, | |
| "expected": "COBALT-7319", | |
| "actual": "COBALT-7319", | |
| "peak_memory_gb": 15.190558278 | |
| } | |
| } | |
| } | |
| }, | |
| "release_status": "experimental-community-family-release", | |
| "variant_path": "4bit" | |
| } | |