Models › Orbit-3.2-Nano

Orbit-3.2-Nano: VRAM needs and speed by GPU

At Q4_K_M, Orbit-3.2-Nano needs about 2.6 GB of VRAM. No measured runs yet. Speeds below are estimates. Submit a run.

2.7B parameters · license: apache-2.0 · 131,072-token context · GGUF published Oct 18, 2026 · GGUF files on Hugging Face

Memory needed per quantization

QuantFile sizeNeeds (estimate)Smallest card it fits
IQ2_M1.0 GB1.9 GB8 GB cards
Q2_K1.1 GB1.9 GB8 GB cards
IQ3_XXS1.2 GB2.1 GB8 GB cards
Q3_K_S1.2 GB2.1 GB8 GB cards
IQ3_XS1.2 GB2.1 GB8 GB cards
Q3_K_M1.3 GB2.2 GB8 GB cards
Q3_K_L1.4 GB2.3 GB8 GB cards
IQ3_M1.4 GB2.3 GB8 GB cards
IQ4_XS1.6 GB2.5 GB8 GB cards
Q4_01.6 GB2.5 GB8 GB cards
Q4_K_S1.7 GB2.5 GB8 GB cards
Q4_11.7 GB2.6 GB8 GB cards
IQ4_NL1.8 GB2.6 GB8 GB cards
Q4_K_M1.8 GB2.6 GB8 GB cards
Q4_K_L1.9 GB2.7 GB8 GB cards
Q5_K_S2.0 GB2.8 GB8 GB cards
Q5_K_M2.1 GB3.0 GB8 GB cards
Q6_K_S2.3 GB3.1 GB8 GB cards
Q6_K2.4 GB3.2 GB8 GB cards
Q6_K_L2.5 GB3.3 GB8 GB cards
Q8_02.9 GB3.7 GB8 GB cards
BF165.4 GB6.3 GB8 GB cards

Expected speed on popular hardware

HardwareBest quant that fitsExpected tok/s
NVIDIA RTX 3060Q8_0107 estimate
NVIDIA RTX 4060 TiQ8_085.2 estimate
NVIDIA RTX 3090Q8_0228 estimate
NVIDIA RTX 4090Q8_0253 estimate
NVIDIA RTX 5090Q8_0399 estimate
AMD Radeon RX 7900 XTXQ8_0254 estimate
Apple M4 Pro (20-core GPU) (64 GB)Q8_080.5 estimate
Apple M3 Max (40-core GPU) (64 GB)Q8_0118 estimate
AMD Ryzen AI Max+ 395 (Radeon 8060S)Q8_072.3 estimate

Check your own card

Too big for your card?

Run Orbit-3.2-Nano on a rented 24-80 GB GPU by the hour.

Rent a bigger GPU on RunPod

Affiliate link: if you sign up, RigBench may earn a referral credit. It never affects rankings or numbers.