Can Ornith 1.5 122B run on the H200?
Yes. At 8K context the best fit is Q6_K, using about 96.6 GB of the 134 GB usable on the H200. Estimated speed ceiling: ~459 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) | 44.7 GB Fits | 44.7 GB Fits | 45.3 GB Fits | 47.6 GB Fits | 1004 |
| 3-bit (Q3_K_M) | 57.9 GB Fits | 58.0 GB Fits | 58.6 GB Fits | 60.8 GB Fits | 770 |
| 4-bit (Q4_K_M) | 72.2 GB Fits | 72.3 GB Fits | 72.8 GB Fits | 75.1 GB Fits | 616 |
| 5-bit (Q5_K_M) | 84.0 GB Fits | 84.1 GB Fits | 84.6 GB Fits | 86.9 GB Fits | 529 |
| 6-bit (Q6_K) | 96.5 GB Fits | 96.6 GB Fits | 97.2 GB Fits | 99.4 GB Fits | 459 |
| 8-bit (Q8_0) • | 124 GB Tight | 125 GB Tight | 125 GB Tight | 127 GB Tight | 355 |
| 16-bit (F16/BF16) • | 233 GB No | 233 GB No | 234 GB No | 236 GB No | — |
• real GGUF file size from unsloth/Ornith-1.5-122B-GGUF. Speed is a bandwidth ceiling (4800 GB/s), not a benchmark. Methodology.
More
Ornith 1.5 122B on every GPU and Mac · Every model that runs on the H200 · Other models that fit 64 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (ornith-ai/Ornith-1.5-122B), NVIDIA specs. Data as of 2026-11-17.