Best home rig for Ornith 1.5 122B

Will it fit my GPU? Best value GPU per euro →

Real, measured generation speeds for Ornith 1.5 122B on home hardware, fastest first. Different quants/engines aren't directly comparable.

Build	Quant	Engine	tok/s	Cost
Ryzen 9 9950X3D + RTX 5090 + 128 GB DDR5: four big MoEs with experts offloaded to RAM (Qwen3.8 Flash-Next 78.8 tok/s)	EXL3 4 bpw, 180/256 experts per layer in RAM, no drafting	TabbyAPI + ExLlamaV3 1.6.2	46.4		source

Rent before you buy: try this model on a cloud GPU for a few dollars first: Runpod (referral link; sign up with Google to get the new-user bonus)

Memory math: how much VRAM / RAM does Ornith 1.5 122B need?

Ornith 1.5 122B has 125.1B parameters, about 10B active per token (source: Hugging Face: 125B incl. vision tower + MTP layer (122B language model, 10B active per halt95/Ornith-1.5-122B-A10B card)). Weight size = parameters × bits-per-weight ÷ 8, before context (KV cache) and runtime overhead.

The source benchmark's EXL3 4 bpw quant is a 65.5 GB download (dev.to, Seppe Gadeyne).

Format	Bits/weight	Weights	Smallest common memory size that holds the weights alone
Q8_0	8.50	132.9 GB	192 GB
Q6_K	6.56	102.6 GB	128 GB
Q5_K_M	5.70	89.2 GB	96 GB
Q4_K_M	4.89	76.5 GB	96 GB
Q3_K_M	4.00	62.5 GB	64 GB
Q2_K	3.16	49.4 GB	64 GB
As measured: EXL3 4 bpw, 180/256 experts per layer in RAM, no drafting	4.00	≈62.5 GB	64 GB

Bits-per-weight for GGUF quants are llama.cpp's own figures (llama.cpp quantize README). "As measured" rows apply the source's stated bits-per-weight to the parameter count; real files differ slightly because some layers stay at higher precision. Try other combinations in the VRAM calculator.

Why it can still run on smaller GPUs: it is a mixture-of-experts model, so only about 10B of its 125.1B parameters are used for each generated token. Runtimes such as llama.cpp (--n-cpu-moe), ExLlamaV3 or Project Maya can keep most expert weights in system RAM and only the shared layers and hot experts in VRAM. Then you need VRAM + system RAM ≥ weights, and speed is limited mainly by RAM bandwidth.

Measured on setups with less GPU memory than the Q4_K_M weights (so part of the model lived in system RAM or was streamed from disk):

RTX 5090 (32 GB) + 128 GB DDR5-5600 (4x32 GB): 46.4 tok/s (EXL3 4 bpw, 180/256 experts per layer in RAM, no drafting)

FAQ

How much VRAM does Ornith 1.5 122B need?

The weights are about 76.5 GB at Q4_K_M (125.1B params × 4.89 bits ÷ 8, before context/KV cache). Add memory for context (KV cache) and the runtime.

What is the fastest measured hardware for Ornith 1.5 122B?

RTX 5090: 46.4 tok/s generation (TabbyAPI + ExLlamaV3 1.6.2, EXL3 4 bpw, 180/256 experts per layer in RAM, no drafting), from https://dev.to/seppegadeyne/four-big-local-llms-on-one-rtx-5090-and-128-gb-of-ram-benchmarked-side-by-side-2h8j.

Which GPUs have measured results for Ornith 1.5 122B?

RTX 5090.

Can it run on… RTX 5090