Skip to main content

Can the RTX 5090 Run Qwen3 30B-A3B?

Runs

Yes. Qwen3 30B-A3B at Q4_K_M is a 18.6 GB download (Hugging Face file listing for unsloth/Qwen3-30B-A3B-GGUF), and the RTX 5090 has 32 GB of VRAM, leaving about 13.4 GB for the KV cache and runtime. Published runs put generation at a median of 234 tokens per second across 1 cited measurement. Fastest cited run: Hardware Corner.

Model file18.6 GBQ4_K_M
VRAM32 GBRTX 5090
Headroom13.4 GBafter weights
Median speed234 tok/s1 cited run
Bandwidth ceiling~892 tok/s1792 GB/s ÷ 2 GB

Published runs for Qwen3 30B-A3B on the RTX 5090

GenerationQuantRuntimeContextSourceDate
234 tok/s q4_K_M llama.cpp 4K Hardware Cornerlabelled "qwen3moe:30b-a3b" 2025-06-01

More about this pair

How much VRAM does Qwen3 30B-A3B need?

The Q4_K_M file is 18.6 GB. Add roughly 1.5 GB for the runtime and an ~8K-token context, so about 20.1 GB of VRAM holds it comfortably. Longer contexts grow the KV cache and need more.

What is the fastest Qwen3 30B-A3B can generate on the RTX 5090?

Qwen3 30B-A3B is a mixture-of-experts model that reads about 2 GB of active weights per token. At the RTX 5090's 1792 GB/s of memory bandwidth, that sets a ceiling near 892 tokens per second. Real runs land below it.

Check another pair

See the full matrix → · All RTX 5090 benchmarks →

Qwen3 30B-A3B on other GPUs

Other models on the RTX 5090

How the checks work

The fit is the Q4_K_M file size against the card's VRAM, with 1.5 GB kept for the runtime and an ~8K context. A published run counts only if the pair fits, the run is Q4-class, its source does not call it an estimate, it is under the bandwidth ceiling above, and it is not one figure pasted across many cards. The whole matrix is downloadable as a CSV under CC BY 4.0.

As an Amazon Associate, SpecPicks earns from qualifying purchases.

This page is editorial synthesis based on publicly available information. No independent first-party benchmarking is reported.