Reference · Coverage
Model families
16 documented families, one canonical format. The engine imports GGUF, ONNX, and the native 1BP, detects the architecture, and picks a kernel path. No config, no glue.
The families
Per-family coverage matrices live in docs/model-families/<family>.md. The set: bitnet-bonsai, deepseek, falcon, gemma, gpt-oss, granite, kimi, laguna, llama, mistral, olmo, phi, qwen, smollm, whisper, zyphra.
1BP format
1BP is the single canonical format; importers exist for GGUF, ONNX, Q4NX, and H1B. The quantization path: Q4NX (4-bit, per-group BF16 scales) to dequantized float32, then re-quantized to symmetric INT8 (per-tensor max-abs) once at startup. Activations are quantized on the fly with a fixed 5.0 / 127.0 scale, empirically validated on Strix Halo.
Zyphra, the engine-tuned family
This is the family the engine was tuned against and the one that powers JARVIS by default.
| Model | Params | 1BP size | Backend(s) | Perf |
|---|---|---|---|---|
| ZAYA1-8B | 8.8B | 6.6 GB | HIP / NPU | 64 tok/s HIP |
| ZAYA1-74B-preview | 74B | 45.8 GB (GGUF) | HIP | 16.7 tok/s |
| ZAYA1-VL-8B | 8.8B | n/a | HIP (vision) | n/a |
| ZR1-1.5B | 1.5B | 781 MB | ZINC / NPU | 26 tok/s ZINC |
| BlackMamba-1.5B | 1.5B | 970 MB | Mamba1 HIP | 79.4 tok/s |
| BlackMamba-2.8B | 2.8B | 1.8 GB | Mamba1 HIP | 46.0 tok/s |
| Zamba2-1.2B-v2 | 1.2B | 1.1 GB | HIP / CPU | 30 tok/s HIP |
| Zamba2-2.7B-v2 | 2.7B | 2.4 GB | HIP / CPU | n/a |
| Zamba2-7B-v2 | 7B | 6.6 GB | HIP / CPU | n/a |
| Zamba-7B-v1 | 7B | 4.3 GB | Mamba1 HIP | n/a |
ZAYA1-8B 1BP is ~6.6 GB full-weight; the 149 MB entry on HF is MoE-expert-stripped. And the 739 MB 1BP for the 74B is physically impossible for a full 74B model; the runnable Q4_K_M GGUF is 45.8 GB.