Can OLMo 2 0425 1B run on the H200?
Yes. At 8K context the best fit is F16, using about 4.4 GB of the 134 GB usable on the H200. Estimated speed ceiling: ~1614 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | Max tok/s (est., 8K) |
|---|---|---|
| 2-bit (Q2_K) • | 1.7 GB Fits | 7579 |
| 3-bit (Q3_K_M) • | 1.8 GB Fits | 6162 |
| 4-bit (Q4_K_M) • | 2.0 GB Fits | 5131 |
| 5-bit (Q5_K_M) • | 2.1 GB Fits | 4467 |
| 6-bit (Q6_K) • | 2.2 GB Fits | 3928 |
| 8-bit (Q8_0) • | 2.6 GB Fits | 3035 |
| 16-bit (F16/BF16) • | 3.9 GB Fits | 1614 |
• real GGUF file size from allenai/OLMo-2-0425-1B-GGUF. Speed is a bandwidth ceiling (4800 GB/s), not a benchmark. Methodology.
More
OLMo 2 0425 1B 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 (allenai/OLMo-2-0425-1B), NVIDIA specs. Data as of 2026-11-17.