Can you run Ornith-1.5-397B locally?
Ornith 1.5 family · 2026 · 403.4B parameters · Hugging Face model card
Ornith-1.5-397B is the flagship of the Ornith 1.5 family: a 397B-class mixture-of-experts open-weight model (HF safetensors.total=403,397,928,944) with 512 routed experts and 10 activated per token, plus a shared expert. The architecture is hybrid: linear-attention layers with full attention every fourth layer. Native context is 262,144 tokens, MIT licensed. At Q4-class on the 403B total that is roughly 256 GB (calculated); the vendor card itself says ~800 GB at bf16 and recommends 8-way tensor parallelism on one node, so this is a multi-GPU server model, not a desktop one. Measured local speeds: not yet published here.
Minimum: 256 GB+ memory (Q4-class on 403B total, calculated) · Recommended: multi-GPU server only (vendor card: 8-way TP, e.g. 8x H200 141GB)
How much VRAM does Ornith-1.5-397B need at each quantization?
Ornith-1.5-397B needs 266.7 GB of VRAM at Q4_K_M. The table below lists weights-only size and total VRAM including overhead for each common quantization level with 4k and 32k token contexts.
| Quantization | Bits / weight | Weights only | Total + KV @4k ctx | Total + KV @32k ctx |
|---|---|---|---|---|
| Q4_K_M | 4.8 | 242.0 GB | 266.7 GB | 270.3 GB |
| Q5_K_M | 5.7 | 287.4 GB | 316.7 GB | 320.2 GB |
| Q6_K | 6.6 | 332.8 GB | 366.6 GB | 370.1 GB |
| Q8_0 | 8.5 | 428.6 GB | 472.0 GB | 475.5 GB |
| FP16 | 16 | 806.8 GB | 888.0 GB | 891.5 GB |
hybrid linear_attention + full_attention every 4th layer (full_attention_interval=4); mrope interleaved
Method: weights = parameters × bits-per-weight (Q4_K_M ≈ 4.8, Q5_K_M ≈ 5.7, Q6_K ≈ 6.6, Q8_0 ≈ 8.5, FP16 = 16), plus 10% loading overhead, plus KV cache from the verified architecture config (60 layers, 2 KV heads, 256 head dim). Source: HF-verified 2026-10-06 via api (gated=false private=false params=403.4B safetensors.total=403397928944 ctx=262144 license=mit repo=ornith-ai/Ornith-1.5-397B@8f6cc8a7aea505364523f84ccf37706e8aea0ee7 layers=60 hidden=4096 heads=32 kv_heads=2 experts=512 per_tok=10). active_params_b=None (dense model; total params used). VRAM figures are documented calculations (Q4-class, 0.6 GB per B params), not measurements. No tok/s invented.
Which GPUs can run Ornith-1.5-397B locally?
At Q4_K_M with a 4k context, Ornith-1.5-397B needs 266.7 GB of VRAM. The lists below are computed live from our GPU database and grouped by how much headroom the card has. Disclosure: CompareAIHardware.com participates in the Amazon Associates program and earns from qualifying purchases through links on this page. Affiliate relationships do not influence our recommendations.
Minimum GPUs that fit Ornith-1.5-397B at Q4_K_M
These GPUs hold the model but leave little headroom — keep contexts short.
| GPU | VRAM | Bandwidth | Type | |
|---|---|---|---|---|
| AMD Instinct MI355X | 288 GB | 8000 GB/s | Data Center GPU | |
| NVIDIA B300 (Blackwell Ultra) | 288 GB | 8000 GB/s | Data Center GPU |
Needs 2+ GPUs to run Ornith-1.5-397B
One of these cards is too small on its own, but a pair (tensor or pipeline parallel, ~90% efficiency) covers the 266.7 GB requirement. See our multi-GPU guide for setup.
| GPU | VRAM | Bandwidth | Type | |
|---|---|---|---|---|
| AMD Instinct MI300X | 192 GB ×2 | 5300 GB/s | Data Center GPU | |
| NVIDIA B200 | 192 GB ×2 | 8000 GB/s | Data Center GPU | |
| AMD Instinct MI325X | 256 GB ×2 | 6000 GB/s | Data Center GPU |
Can a Mac run Ornith-1.5-397B?
Yes — these Apple Silicon machines fit Ornith-1.5-397B at Q4_K_M, since macOS lets the GPU use about 75% of unified memory. Generation speed is bound by memory bandwidth, so the GB/s column matters as much as capacity.
| Apple Silicon | Unified memory | Usable by GPU (~75%) | Bandwidth |
|---|---|---|---|
| M5 Ultra (Mac Studio) | 512 GB | 384 GB | 1200 GB/s |