NVIDIA RTX 4090

Ada Lovelace · 2022

VRAM

24 GB

Bandwidth

1008 GB/s

MSRP

$1,599

Models

21 compatible

Compatible Models (21)

Qwen 3.5 27B

Alibaba · 27B

Q4_K_M

✓ Full quality

Qwen 3.5 3B

Alibaba · 3B

Q8_0

✓ Full quality

Qwen 3.5 9B

Alibaba · 9B

Q8_0

✓ Full quality

Qwen 3.6 27B

Alibaba · 27B

Q4_K_M

✓ Full quality

Qwen 3.6 35B-A3B

Alibaba · 3B active / 35B total (MoE)

Q4_K_M

✓ Full quality

DeepSeek R1 7B

DeepSeek · 7B

Q8_0

✓ Full quality

Gemma 4 27B

Google · 27B (4B active MoE)

Q4_K_M

✓ Full quality

Gemma 4 31B

Google · 31B

Q4_K_M

✓ Full quality

Gemma 4 E2B

Google · 2.3B active / 5B total

Q8_0

✓ Full quality

Gemma 4 E4B

Google · 4B active (MoE)

Q8_0

✓ Full quality

Gemma 4 26B-A4B

Google DeepMind · 26B (4B active MoE)

Q4_K_M

✓ Full quality

micro-kiki v3

L'Électron Rare · 3B active / 35B total (MoE + 35 LoRAs)

Q4_K_M

✓ Full quality

Phi-4

Microsoft · 14B

Q8_0

✓ Full quality

Mistral Small 3.2

Mistral AI · 22B

Q4_K_M

✓ Full quality

Nemotron-3 Nano Omni 30B-A3B

NVIDIA · 3.5B active / 30B total (Mamba-2 + MoE + Attention)

Q4_K_M

✓ Full quality

MiniCPM 4.6

OpenBMB · 1.2B

Q8_0

✓ Full quality

Poolside Laguna XS.2

Poolside AI · 3B active / 33B total (MoE)

Q4_K_M

✓ Full quality

Ternary Bonsai 1.7B

PrismML · 1.7B

1.58-bit (MLX)

✓ Full quality

Ternary Bonsai 4B

PrismML · 4B

1.58-bit (MLX)

✓ Full quality

Ternary Bonsai 8B

PrismML · 8B

1.58-bit (MLX)

✓ Full quality

Zyphra ZAYA1-8B

Zyphra AI · 8B (MoE)

Q8_0

✓ Full quality

Requires more VRAM (6)

Qwen 3.5 72B

Alibaba

Needs 27GB+

Llama 4 Scout