· Hardware

RTX 3090

NVIDIA Ampere · validated on NVIDIA RTX 3090 · 12 models in production · last verified Sep 9, 2026.

ModelPrecisionCapabilitiesSinceLast verifiedBenchmark
Gemma 4 E2BGemmaBF16Query · Captionv2.2Sep 9, 2026
Gemma 4 E4BGemmaBF16Query · Captionv2.2Sep 9, 2026
Moondream 2Moondream VLMBF16Query · Caption · Detect · Point · Segmentv2.2Sep 9, 2026
Moondream 3Moondream VLMBF16Query · Caption · Detect · Point · Segmentv2.2Sep 9, 2026
Parakeet TDT 0.6B v3ParakeetBF16Transcribev2.2Sep 9, 2026
Qwen 3.5 0.8BQwen 3BF16Query · Captionv2.2Sep 9, 2026
Qwen 3.5 2BQwen 3BF16Query · Captionv2.2Sep 9, 2026
Qwen 3.5 4BQwen 3BF16Query · Captionv2.2Sep 9, 2026
Qwen 3.5 9BQwen 3BF16Query · Captionv2.2Sep 9, 2026
Qwen3-ASR 0.6BQwen3-ASRBF16Transcribev2.2Sep 9, 2026
Qwen3-ASR 1.7BQwen3-ASRBF16Transcribev2.2Sep 9, 2026
Whisper large-v3-turboWhisperBF16Transcribev2.2Sep 9, 2026
Run on RTX 3090

Install the SDK and run any listed model locally.

$pip install moondream

Quickstart docs · all architectures · product updates