Image-Text-to-Text
MLX
Safetensors
qwen3_5
mtplx
qwen3.6
speculative-decoding
multi-token-prediction
5-bit
conversational
Instructions to use deanzchen/Qwen3.6-27B-Q5-MTPLX with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use deanzchen/Qwen3.6-27B-Q5-MTPLX 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("deanzchen/Qwen3.6-27B-Q5-MTPLX") config = load_config("deanzchen/Qwen3.6-27B-Q5-MTPLX") # 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
- Pi
How to use deanzchen/Qwen3.6-27B-Q5-MTPLX with Pi:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "deanzchen/Qwen3.6-27B-Q5-MTPLX"
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "mlx-lm": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "deanzchen/Qwen3.6-27B-Q5-MTPLX" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Hermes Agent
How to use deanzchen/Qwen3.6-27B-Q5-MTPLX with Hermes Agent:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "deanzchen/Qwen3.6-27B-Q5-MTPLX"
Configure Hermes
# Install Hermes: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash hermes setup # Point Hermes at the local server: hermes config set model.provider custom hermes config set model.base_url http://127.0.0.1:8080/v1 hermes config set model.default deanzchen/Qwen3.6-27B-Q5-MTPLX
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use deanzchen/Qwen3.6-27B-Q5-MTPLX with OpenClaw:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "deanzchen/Qwen3.6-27B-Q5-MTPLX"
Configure OpenClaw
# Install OpenClaw: npm install -g openclaw@latest # Register the local server and set it as the default model: openclaw onboard --non-interactive --mode local \ --auth-choice custom-api-key \ --custom-base-url http://127.0.0.1:8080/v1 \ --custom-model-id "deanzchen/Qwen3.6-27B-Q5-MTPLX" \ --custom-provider-id mlx-lm \ --custom-compatibility openai \ --custom-text-input \ --accept-risk \ --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
| { | |
| "schema_version": 1, | |
| "artifact": "deanzchen/Qwen3.6-27B-Q5-MTPLX", | |
| "hardware": { | |
| "chip": "Apple M5 Max", | |
| "gpu_cores": 40, | |
| "unified_memory_gib": 48, | |
| "power_mode_for_latest_check": "high_power" | |
| }, | |
| "runtime": { | |
| "mtplx": "2.0.2", | |
| "mlx": "0.31.2", | |
| "mlx_lm": "0.31.3", | |
| "profile": "sustained", | |
| "generation_mode": "mtp", | |
| "depth": 2, | |
| "context_window": 262144, | |
| "prefill_chunk_tokens": 512, | |
| "trunk_quantization": "5-bit affine group64", | |
| "draft_head_quantization": "4-bit affine group64", | |
| "target_sampler": { | |
| "temperature": 0.6, | |
| "top_p": 0.95, | |
| "top_k": 20 | |
| }, | |
| "draft_sampler_for_ui_measurements": { | |
| "temperature": 0.6, | |
| "top_p": 0.95, | |
| "top_k": 20 | |
| }, | |
| "paged_kv_quantization": "off", | |
| "turboquant_note": "K8/V4 affects extreme-context cache capacity, not these short decode measurements." | |
| }, | |
| "verification_schedule": { | |
| "defer_verify_hidden_eval": true, | |
| "verify_hidden_mode": "logits_first_committed_slice", | |
| "exact_token_identity_verified": true | |
| }, | |
| "expected_decode_tok_s": { | |
| "direct_tuner_d2": { | |
| "expected_range": [45.0, 50.0], | |
| "warning_below": 45.0, | |
| "failure_below": 42.0, | |
| "workload": "mtplx long_code, thinking off, seed 0, 128 or 256 generated tokens" | |
| }, | |
| "direct_tuner_d2_high_power": { | |
| "expected_range": [45.0, 50.0], | |
| "warning_below": 45.0, | |
| "workload": "mtplx long_code, thinking off, seed 0, 128 generated tokens" | |
| }, | |
| "openai_api_code_first": { | |
| "expected_range": [40.0, 44.0], | |
| "requirement": "Send X-MTPLX-Allow-Client-Controls: 1 with enable_thinking=false." | |
| }, | |
| "eight_token_smoke": { | |
| "expected_range": [35.0, 46.0], | |
| "regression_gate": false, | |
| "note": "Too short for stable throughput measurement." | |
| } | |
| }, | |
| "measurements": { | |
| "historical_setup_128_d2": 48.799728436870346, | |
| "historical_comparison_256_d2": 49.90875607132132, | |
| "hub_direct_128_d2": 43.92720981573101, | |
| "hub_high_power_direct_128_d2": 45.419032271264705, | |
| "hub_post_reboot_high_power_direct_128_d2": 46.27235162628437, | |
| "hub_post_reboot_high_power_direct_128_d2_variance_5x": { | |
| "runs": [ | |
| 46.47503852458389, | |
| 46.38583666459772, | |
| 46.33124692797028, | |
| 46.27492432091384, | |
| 46.316507149484536 | |
| ], | |
| "mean": 46.356710717510055, | |
| "median": 46.33124692797028, | |
| "sample_standard_deviation": 0.07714197438579953, | |
| "coefficient_of_variation_percent": 0.16640950833610618, | |
| "minimum": 46.27492432091384, | |
| "maximum": 46.47503852458389, | |
| "range": 0.20011420367004717, | |
| "all_runs_above_45": true, | |
| "acceptance_rate_all_runs": 1.0 | |
| }, | |
| "original_local_deferred_hidden_direct_128_d2_variance_5x": { | |
| "runs": [ | |
| 50.67376896277482, | |
| 50.9484927857385, | |
| 50.75530318826678, | |
| 50.7859195241367, | |
| 50.750990759454915 | |
| ], | |
| "mean": 50.782895044074344, | |
| "median": 50.75530318826678, | |
| "sample_standard_deviation": 0.10138966296877788, | |
| "coefficient_of_variation_percent": 0.19965317629249388, | |
| "minimum": 50.67376896277482, | |
| "maximum": 50.9484927857385, | |
| "all_runs_at_or_above_48_8": true, | |
| "acceptance_rate_all_runs": 1.0, | |
| "token_sha256": "f875ce74f006ba502dac263111e6c9b1b1b4bca03648c43f670c8cae8e0a58af" | |
| }, | |
| "persisted_tune_wrapper_final_gate": { | |
| "decode_tok_s": 49.81852161488856, | |
| "acceptance_rate": 1.0, | |
| "accepted_by_depth": [43, 42], | |
| "drafted_by_depth": [43, 42], | |
| "token_sha256": "f875ce74f006ba502dac263111e6c9b1b1b4bca03648c43f670c8cae8e0a58af", | |
| "text_sha256": "abc2ec0ce79cbf90e38043eeacd0fc9d34d624780e6cc630b3d5fbd0e3ced05e", | |
| "target_48_8_met": true | |
| }, | |
| "hub_direct_256_d2_runs": [43.830748328265656, 43.695076417584225], | |
| "hub_api_code_first_three_run_mean": 41.46785703207608, | |
| "hub_api_auto_reasoning_three_run_mean": 39.68670477257572, | |
| "latest_ar_128": 24.149735339889887, | |
| "latest_d2_speedup_vs_ar": 1.916060402941681, | |
| "high_power_improvement_percent": 5.338699681566217, | |
| "post_reboot_improvement_percent": 1.878770445665201 | |
| }, | |
| "validation": { | |
| "status": "pass", | |
| "model_files_match_original_sha256": true, | |
| "direct_128_accepted_by_depth": [43, 42], | |
| "direct_128_drafted_by_depth": [43, 42], | |
| "direct_256_acceptance_rate": 0.9883720930232558, | |
| "conclusion": "No artifact or MTP acceptance regression. Current throughput is within the expected machine-state range." | |
| } | |
| } | |