| {"variant_id":"prism-ml/Bonsai-1.7B-unpacked","family":"Bonsai","source_model_id":"Qwen/Qwen3-1.7B","category":"llm","total_parameters_b":1.7,"format":"safetensors-fp16","bits_per_weight":16,"memory_estimate_gb":3.4,"memory_estimate_basis":"derived-weight-only","fit_status":"yes","license":"apache-2.0","license_status":"needs-card-review","official_checkpoint":true,"runtime_support":["transformers"],"source_url":"https://huggingface.co/prism-ml/Bonsai-1.7B-unpacked","last_verified":"2026-07-15","notes":"Official Prism ML unpacked checkpoint; retain Qwen3 base parameter identity and do not count this as a new base family."} | |
| {"variant_id":"prism-ml/Bonsai-4B-unpacked","family":"Bonsai","source_model_id":"Qwen/Qwen3-4B","category":"llm","total_parameters_b":4.0,"format":"safetensors-fp16","bits_per_weight":16,"memory_estimate_gb":8.0,"memory_estimate_basis":"derived-weight-only","fit_status":"yes","license":"apache-2.0","license_status":"needs-card-review","official_checkpoint":true,"runtime_support":["transformers"],"source_url":"https://huggingface.co/prism-ml/Bonsai-4B-unpacked","last_verified":"2026-07-15","notes":"Official Prism ML unpacked checkpoint; retain Qwen3 base parameter identity and do not count this as a new base family."} | |
| {"variant_id":"prism-ml/Bonsai-8B-unpacked","family":"Bonsai","source_model_id":"Qwen/Qwen3-8B","category":"llm","total_parameters_b":8.0,"format":"safetensors-fp16","bits_per_weight":16,"memory_estimate_gb":16.0,"memory_estimate_basis":"derived-weight-only","fit_status":"borderline","license":"apache-2.0","license_status":"needs-card-review","official_checkpoint":true,"runtime_support":["transformers"],"source_url":"https://huggingface.co/prism-ml/Bonsai-8B-unpacked","last_verified":"2026-07-15","notes":"Weight-only estimate consumes the full 16GB budget before runtime overhead; not recommended for a 16GB device without quantization."} | |
| {"variant_id":"prism-ml/Ternary-Bonsai-8B-unpacked","family":"Ternary Bonsai","source_model_id":"Qwen/Qwen3-8B","category":"llm","total_parameters_b":8.0,"format":"safetensors-fp16-unpacked","bits_per_weight":16,"memory_estimate_gb":16.0,"memory_estimate_basis":"official-card-weight-only","fit_status":"borderline","license":"apache-2.0","license_status":"verified","official_checkpoint":true,"runtime_support":["transformers"],"source_url":"https://huggingface.co/prism-ml/Ternary-Bonsai-8B-unpacked","last_verified":"2026-07-15","notes":"The official card describes this as an unpacked FP16 file and explicitly says the packed format is where the memory benefit comes from."} | |
| {"variant_id":"prism-ml/Bonsai-1.7B-mlx-1bit","family":"Bonsai","source_model_id":"Qwen/Qwen3-1.7B","category":"llm","total_parameters_b":1.7,"format":"mlx-1bit","bits_per_weight":1.25,"memory_estimate_gb":0.27,"memory_estimate_basis":"derived-packed-weight-estimate","fit_status":"yes","license":"apache-2.0","license_status":"needs-card-review","official_checkpoint":true,"runtime_support":["mlx"],"source_url":"https://huggingface.co/prism-ml/Bonsai-1.7B-mlx-1bit","last_verified":"2026-07-15","notes":"Packed Apple Silicon deployment variant; total parameters remain 1.7B despite the compressed footprint."} | |
| {"variant_id":"prism-ml/Bonsai-4B-mlx-1bit","family":"Bonsai","source_model_id":"Qwen/Qwen3-4B","category":"llm","total_parameters_b":4.0,"format":"mlx-1bit","bits_per_weight":1.25,"memory_estimate_gb":0.63,"memory_estimate_basis":"official-card-weight-only","fit_status":"yes","license":"apache-2.0","license_status":"verified","official_checkpoint":true,"runtime_support":["mlx"],"source_url":"https://huggingface.co/prism-ml/Bonsai-4B-mlx-1bit","last_verified":"2026-07-15","notes":"Official card reports a 0.63GB deployed weight footprint and Apple Silicon/iPhone support."} | |
| {"variant_id":"prism-ml/Bonsai-8B-mlx-1bit","family":"Bonsai","source_model_id":"Qwen/Qwen3-8B","category":"llm","total_parameters_b":8.0,"format":"mlx-1bit","bits_per_weight":1.25,"memory_estimate_gb":1.26,"memory_estimate_basis":"derived-packed-weight-estimate","fit_status":"yes","license":"apache-2.0","license_status":"needs-card-review","official_checkpoint":true,"runtime_support":["mlx"],"source_url":"https://huggingface.co/prism-ml/Bonsai-8B-mlx-1bit","last_verified":"2026-07-15","notes":"Packed Apple Silicon deployment variant; total parameters remain 8B despite the compressed footprint."} | |
| {"variant_id":"prism-ml/Ternary-Bonsai-1.7B-mlx-2bit","family":"Ternary Bonsai","source_model_id":"Qwen/Qwen3-1.7B","category":"llm","total_parameters_b":1.7,"format":"mlx-2bit","bits_per_weight":2.0,"memory_estimate_gb":0.43,"memory_estimate_basis":"derived-packed-weight-estimate","fit_status":"yes","license":"apache-2.0","license_status":"needs-card-review","official_checkpoint":true,"runtime_support":["mlx"],"source_url":"https://huggingface.co/prism-ml/Ternary-Bonsai-1.7B-mlx-2bit","last_verified":"2026-07-15","notes":"Packed ternary deployment variant; not a separate base model."} | |
| {"variant_id":"prism-ml/Ternary-Bonsai-4B-mlx-2bit","family":"Ternary Bonsai","source_model_id":"Qwen/Qwen3-4B","category":"llm","total_parameters_b":4.0,"format":"mlx-2bit","bits_per_weight":2.0,"memory_estimate_gb":1.0,"memory_estimate_basis":"derived-packed-weight-estimate","fit_status":"yes","license":"apache-2.0","license_status":"needs-card-review","official_checkpoint":true,"runtime_support":["mlx"],"source_url":"https://huggingface.co/prism-ml/Ternary-Bonsai-4B-mlx-2bit","last_verified":"2026-07-15","notes":"Packed ternary deployment variant; not a separate base model."} | |
| {"variant_id":"prism-ml/Ternary-Bonsai-8B-mlx-2bit","family":"Ternary Bonsai","source_model_id":"Qwen/Qwen3-8B","category":"llm","total_parameters_b":8.0,"format":"mlx-2bit","bits_per_weight":2.0,"memory_estimate_gb":2.0,"memory_estimate_basis":"derived-packed-weight-estimate","fit_status":"yes","license":"apache-2.0","license_status":"needs-card-review","official_checkpoint":true,"runtime_support":["mlx"],"source_url":"https://huggingface.co/prism-ml/Ternary-Bonsai-8B-mlx-2bit","last_verified":"2026-07-15","notes":"Packed ternary deployment variant; not a separate base model."} | |
| {"variant_id":"prism-ml/Bonsai-1.7B-gguf","family":"Bonsai","source_model_id":"Qwen/Qwen3-1.7B","category":"llm","total_parameters_b":1.7,"format":"gguf","bits_per_weight":null,"memory_estimate_gb":null,"memory_estimate_basis":"requires-quantization-level","fit_status":"unknown","license":"apache-2.0","license_status":"needs-card-review","official_checkpoint":true,"runtime_support":["llama_cpp"],"source_url":"https://huggingface.co/prism-ml/Bonsai-1.7B-gguf","last_verified":"2026-07-15","notes":"GGUF family record; exact memory depends on the selected quantization file."} | |
| {"variant_id":"prism-ml/Bonsai-4B-gguf","family":"Bonsai","source_model_id":"Qwen/Qwen3-4B","category":"llm","total_parameters_b":4.0,"format":"gguf","bits_per_weight":null,"memory_estimate_gb":null,"memory_estimate_basis":"requires-quantization-level","fit_status":"unknown","license":"apache-2.0","license_status":"needs-card-review","official_checkpoint":true,"runtime_support":["llama_cpp"],"source_url":"https://huggingface.co/prism-ml/Bonsai-4B-gguf","last_verified":"2026-07-15","notes":"GGUF family record; exact memory depends on the selected quantization file."} | |
| {"variant_id":"prism-ml/Bonsai-8B-gguf","family":"Bonsai","source_model_id":"Qwen/Qwen3-8B","category":"llm","total_parameters_b":8.0,"format":"gguf","bits_per_weight":null,"memory_estimate_gb":null,"memory_estimate_basis":"requires-quantization-level","fit_status":"unknown","license":"apache-2.0","license_status":"needs-card-review","official_checkpoint":true,"runtime_support":["llama_cpp"],"source_url":"https://huggingface.co/prism-ml/Bonsai-8B-gguf","last_verified":"2026-07-15","notes":"GGUF family record; exact memory depends on the selected quantization file."} | |
| {"variant_id":"prism-ml/bonsai-image-binary-4B-unpacked","family":"Bonsai Image","source_model_id":null,"category":"image-generation","total_parameters_b":4.0,"format":"safetensors-fp16","bits_per_weight":16,"memory_estimate_gb":8.0,"memory_estimate_basis":"derived-weight-only","fit_status":"yes","license":"apache-2.0","license_status":"needs-card-review","official_checkpoint":true,"runtime_support":["mlx"],"source_url":"https://huggingface.co/prism-ml/bonsai-image-binary-4B-unpacked","last_verified":"2026-07-15","notes":"Image-generation deployment variant; kept outside the language-model count."} | |
| {"variant_id":"prism-ml/bonsai-image-ternary-4B-unpacked","family":"Bonsai Image","source_model_id":null,"category":"image-generation","total_parameters_b":4.0,"format":"safetensors-fp16-unpacked","bits_per_weight":16,"memory_estimate_gb":8.0,"memory_estimate_basis":"derived-weight-only","fit_status":"yes","license":"apache-2.0","license_status":"needs-card-review","official_checkpoint":true,"runtime_support":["mlx"],"source_url":"https://huggingface.co/prism-ml/bonsai-image-ternary-4B-unpacked","last_verified":"2026-07-15","notes":"Image-generation deployment variant; kept outside the language-model count."} | |
| {"variant_id":"prism-ml/bonsai-image-binary-4B-mlx-1bit","family":"Bonsai Image","source_model_id":null,"category":"image-generation","total_parameters_b":4.0,"format":"mlx-1bit","bits_per_weight":1.25,"memory_estimate_gb":0.63,"memory_estimate_basis":"derived-packed-weight-estimate","fit_status":"yes","license":"apache-2.0","license_status":"needs-card-review","official_checkpoint":true,"runtime_support":["mlx"],"source_url":"https://huggingface.co/prism-ml/bonsai-image-binary-4B-mlx-1bit","last_verified":"2026-07-15","notes":"Packed image-generation variant; kept outside the language-model count."} | |
| {"variant_id":"prism-ml/bonsai-image-ternary-4B-mlx-2bit","family":"Bonsai Image","source_model_id":null,"category":"image-generation","total_parameters_b":4.0,"format":"mlx-2bit","bits_per_weight":2.0,"memory_estimate_gb":1.0,"memory_estimate_basis":"derived-packed-weight-estimate","fit_status":"yes","license":"apache-2.0","license_status":"needs-card-review","official_checkpoint":true,"runtime_support":["mlx"],"source_url":"https://huggingface.co/prism-ml/bonsai-image-ternary-4B-mlx-2bit","last_verified":"2026-07-15","notes":"Packed image-generation variant; kept outside the language-model count."} | |
| {"variant_id":"prism-ml/Bonsai-27B-mlx-1bit","family":"Bonsai","source_model_id":"Qwen/Qwen3.6-27B","category":"llm","parameter_scope":"edge-exception","total_parameters_b":27.76,"format":"mlx-1bit","bits_per_weight":1.25,"memory_estimate_gb":12.2,"memory_estimate_basis":"official-card-peak-at-100k-context","fit_status":"yes","license":"apache-2.0","license_status":"verified","official_checkpoint":true,"runtime_support":["mlx","mlx-swift"],"source_url":"https://huggingface.co/prism-ml/Bonsai-27B-mlx-1bit","last_verified":"2026-07-15","notes":"27.3B binary language weights plus a 0.46B vision tower. Official MLX peak is 12.2GB at 100K context; this is included as an edge-deployment exception and not in the primary sub-12B count."} | |
| {"variant_id":"prism-ml/Bonsai-27B-gguf","family":"Bonsai","source_model_id":"Qwen/Qwen3.6-27B","category":"llm","parameter_scope":"edge-exception","total_parameters_b":27.76,"format":"gguf-q1_0_g128","bits_per_weight":1.125,"memory_estimate_gb":11.6,"memory_estimate_basis":"official-card-peak-at-100k-context","fit_status":"yes","license":"apache-2.0","license_status":"verified","official_checkpoint":true,"runtime_support":["llama_cpp"],"source_url":"https://huggingface.co/prism-ml/Bonsai-27B-gguf","last_verified":"2026-07-15","notes":"Official 1-bit GGUF artifact with custom llama.cpp kernels. The card reports 11.6GB peak at 100K context; this is an edge-deployment exception and not in the primary sub-12B count."} | |
| {"variant_id":"prism-ml/Ternary-Bonsai-27B-mlx-2bit","family":"Ternary Bonsai","source_model_id":"Qwen/Qwen3.6-27B","category":"llm","parameter_scope":"edge-exception","total_parameters_b":27.76,"format":"mlx-2bit","bits_per_weight":2.25,"memory_estimate_gb":15.5,"memory_estimate_basis":"official-card-peak-at-100k-context","fit_status":"yes","license":"apache-2.0","license_status":"verified","official_checkpoint":true,"runtime_support":["mlx"],"source_url":"https://huggingface.co/prism-ml/Ternary-Bonsai-27B-mlx-2bit","last_verified":"2026-07-15","notes":"27.3B ternary language weights plus a 0.46B vision tower. The MLX pack uses 2.25 effective bits/weight and the official card reports 15.5GB peak at 100K context; this is an edge-deployment exception and not in the primary sub-12B count."} | |
| {"variant_id":"prism-ml/Ternary-Bonsai-27B-gguf","family":"Ternary Bonsai","source_model_id":"Qwen/Qwen3.6-27B","category":"llm","parameter_scope":"edge-exception","total_parameters_b":27.76,"format":"gguf-q2_0_g128","bits_per_weight":1.71,"memory_estimate_gb":14.7,"memory_estimate_basis":"official-card-peak-at-100k-context","fit_status":"yes","license":"apache-2.0","license_status":"verified","official_checkpoint":true,"runtime_support":["llama_cpp"],"source_url":"https://huggingface.co/prism-ml/Ternary-Bonsai-27B-gguf","last_verified":"2026-07-15","notes":"Official ternary GGUF artifact with custom llama.cpp kernels. The card reports 14.7GB peak at 100K context; this is an edge-deployment exception and not in the primary sub-12B count."} | |