Only with CPU offload

Ornith 1.5 9B at Q4_K_M needs 6.1 GB but only 5.3 GB is addressable, so about 16% of the layers would stream from system RAM at roughly 60 GB/s. Expect around 22 tokens per second — usable for batch work, painful for chat.

84% on GPU 8K context Q4_K_M · 5.3 GB MIT Released 19 Aug 2026

The small Ornith: a coding-agent reasoning build on the Qwen3.5 9B architecture. Same VRAM as its base, thinks before every answer.

Hardware requirements What hardware do I need to run Ornith 1.5 9B? →

At Q4_K_M with an 8K context: a 10 GB GPU (Arc B570) or a 16 GB Mac (M4). Minimum, recommended, high-quality and long-context tiers, with all 114 devices checked at every quantisation.

The VRAM budget

weights 5.3 GB
Weights 5.3 GB KV cache @ 8K 0.25 GB Runtime overhead 0.6 GB Over budget 0.8 GB past 5.3 GB

Quantisation ladder

QuantWeightsTotal @ 8KMax contextTok/sQualityFit
F16 17.5 GB 18.4 GB — ~2.6 Reference 13.1 GB over
Q8_0 9.3 GB 10.2 GB — ~6.4 −0.1% ppl 4.9 GB over
Q6_K 7.2 GB 8.0 GB — ~10 −0.4% ppl 2.7 GB over
Q5_K_M 6.2 GB 7.1 GB — ~14 −0.8% ppl 1.8 GB over
Q4_K_M 5.3 GB 6.1 GB — ~22 −1.9% ppl 0.8 GB over
Q3_K_M 4.3 GB 5.1 GB 13K 47 −5.4% ppl Fits
Q2_K 3.7 GB 4.5 GB 32K 55 −15% ppl Long context

Quality is the published perplexity delta against f16 weights. Max context assumes an f16 KV cache; q8_0 roughly doubles it. Only 8 of its 32 blocks keep a per-token KV cache; the rest are linear-attention, Mamba or convolution blocks with a fixed-size state.

How to run it

terminal
$ llama-server \
    -hf ornith-ai/Ornith-1.5-9B:Q4_K_M \
    -c 8192 -ngl 26

The engine underneath most of the others. Every knob is exposed. More on llama.cpp.

01Download is 5.3 GB. Keep it on an SSD — a first load off a spinning disk takes minutes.
02Close anything else holding VRAM. A browser with hardware acceleration can sit on 1–2 GB.
03If it falls back to CPU silently, drop the context first, then step down a quantisation.
See all models for this rig Compare with another model

Related models on a GeForce GTX 1660 Super

ModelTotalTok/s est.
Ornith 1.5 35B-A3B
35.9B MoE · same family
Q4_K_M · 20.9 GB ~9.2 CPU offload
Ornith 1.5 397B-A17B
397B MoE · same family
Q4_K_M · 224.1 GB — Won't fit
Qwen3.5 9B
9.65B · similar size
Q4_K_M · 6.3 GB ~21 CPU offload
Ministral 3 8B
8.92B · similar size
Q4_K_M · 6.7 GB ~18 CPU offload
Granite 4.1 8B
8.79B · similar size
Q4_K_M · 6.8 GB ~17 CPU offload
LFM2.5 8B-A1B
8.47B MoE · similar size
Q4_K_M · 5.5 GB ~81 CPU offload
Fara 7B
8.29B · similar size
Q4_K_M · 5.7 GB ~31 CPU offload

Every model against the GeForce GTX 1660 Super →

Ornith 1.5 9B on other hardware

DeviceTotalTok/s est.
GeForce RTX 4060 Laptop
7.0 GB usable · 256 GB/s
6.1 GB 29 Tight fit
Arc B570
8.8 GB usable · 380 GB/s
6.1 GB 43 Runs great
GeForce RTX 3060 12 GB
10.6 GB usable · 360 GB/s
6.1 GB 41 Runs great
M4 · 16 GB
10.7 GB usable · 120 GB/s
6.1 GB 13 Runs great
GeForce RTX 4090
22.4 GB usable · 1008 GB/s
6.1 GB 115 Runs great
Radeon RX 7900 XTX
22.4 GB usable · 960 GB/s
6.1 GB 110 Runs great
M3 Max · 64 GB
48.0 GB usable · 400 GB/s
6.1 GB 42 Runs great

All 114 devices, by tier →