Can I run Qwen3 30B-A3B (2507) on the GeForce RTX 5090?

Runs greatQ4_K_M fits in VRAM with 11.5 GB to spare: Runs great, est. 235.0 tok/s (188.0–282.0, calibrated estimate ±20%).

32 GB VRAM · 1,792.0 GB/s · FP16 104.8 TFLOPS. Assumes 32 GB of DDR5-5600 system RAM, Windows with this GPU driving the display, 8k context and an f16 KV cache.

Every quant of Qwen3 30B-A3B (2507) on the GeForce RTX 5090

4 tracked GGUF files, smallest first, at 8k context
QuantFile sizeVerdictSpeedMemoryRuns asNotes
Q2_K11.26 GBRuns great
est. 235.0 tok/s188.0–282.0calibrated estimate ±20%
13.2 / 32.0 GBFull GPU—
IQ4_XS16.38 GBRuns great
est. 235.0 tok/s188.0–282.0calibrated estimate ±20%
18.4 / 32.0 GBFull GPU—
Q4_K_Mbaseline18.56 GBRuns great
est. 235.0 tok/s188.0–282.0calibrated estimate ±20%
20.5 / 32.0 GBFull GPU—
Q8_032.48 GBWon't run—Needs about 33.9 GB; 31.4 GB of VRAM and 28.0 GB of RAM are free——

Where the memory goes at Q4_K_M

Weights, KV cache and compute buffer are the model; the OS reservation is the display driver.
GPU memory
Weights
18.6 GB
KV cache
0.8 GB
Compute buffer
0.6 GB
OS reserve
0.6 GB
Free
11.5 GB

Context length vs. KV cache

KV cache

Memory needed and verdict at Q4_K_M for each context length and KV cache type, with estimated tokens per second.
ContextKV f16KV q8_0KV q4_0
4k
19.5 GBRuns great
est. 235.0 tok/s188.0–282.0calibrated estimate ±20%
19.3 GBRuns great
est. 235.0 tok/s188.0–282.0calibrated estimate ±20%
19.2 GBRuns great
est. 235.0 tok/s188.0–282.0calibrated estimate ±20%
8k
19.9 GBRuns great
est. 235.0 tok/s188.0–282.0calibrated estimate ±20%
19.6 GBRuns great
est. 235.0 tok/s188.0–282.0calibrated estimate ±20%
19.4 GBRuns great
est. 235.0 tok/s188.0–282.0calibrated estimate ±20%
16k
20.8 GBRuns great
est. 222.8 tok/s178.3–267.4calibrated estimate ±20%
20.1 GBRuns great
est. 235.0 tok/s188.0–282.0calibrated estimate ±20%
19.7 GBRuns great
est. 235.0 tok/s188.0–282.0calibrated estimate ±20%
32k
22.6 GBRuns great
est. 154.2 tok/s123.4–185.0calibrated estimate ±20%
21.1 GBRuns great
est. 216.1 tok/s172.9–259.3calibrated estimate ±20%
20.3 GBRuns great
est. 235.0 tok/s188.0–282.0calibrated estimate ±20%
64k
26.1 GBRuns great
est. 95.4 tok/s76.3–114.5calibrated estimate ±20%
23.1 GBRuns great
est. 147.8 tok/s118.3–177.4calibrated estimate ±20%
21.5 GBRuns great
est. 209.8 tok/s167.8–251.7calibrated estimate ±20%
128k
33.1 GBRuns slowly
est. 19.2 tok/s13.4–24.9theoretical estimate ±30%
27.2 GBRuns great
est. 90.6 tok/s72.5–108.7calibrated estimate ±20%
23.9 GBRuns great
est. 142.0 tok/s113.6–170.4calibrated estimate ±20%
256k
47.2 GBRuns slowly
est. 19.2 tok/s13.4–24.9theoretical estimate ±30%
35.2 GBRuns slowly
est. 19.2 tok/s13.4–24.9theoretical estimate ±30%
28.8 GBRuns great
est. 86.2 tok/s69.0–103.5calibrated estimate ±20%

S = Runs great · A = Runs well · B = Runs slowly · F = Won't run

At Q4_K_M the verdict stays Runs great up to 64k context with an f16 KV cache, and up to 128k with q8_0.

Estimated speed

Generation
est. 235.0 tok/s (188.0–282.0, calibrated estimate ±20%)
Prompt processing
about 14,672 tok/s (8,803–20,541, rough estimate ±40%)

Measured on this exact combination

Public benchmarks we calibrate against. Their conditions (context, backend, flags) can differ from the estimates above.

Measured on this exact combination
HardwareQuantBackendContextPrompt (tok/s)Generation (tok/s)FlagsSourceMeasured
GeForce RTX 5090Q4_K_XLllama.cpp16k—141.6—hardware-corner.net2026-08-09

If this is not enough

How to run it

Commands for the Q4_K_M file. Both tools download from Hugging Face on first run.

File: Qwen3-30B-A3B-Instruct-2507-Q4_K_M.gguf (18.56 GB) from unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF.

llama.cpp
llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_M -c 8192 -ngl all -fa on
  • -ngl all loads every layer on the GPU.
  • -fa on enables flash attention (needed for KV cache quantization).
  • --cache-type-k q8_0 --cache-type-v q8_0 shrinks the KV cache to 0.43 GB at 8k context.
Ollama

PowerShell (quit the Ollama tray app first):

$env:OLLAMA_CONTEXT_LENGTH="8192"; ollama serve
ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_M
  • Ollama defaults to 4k context on GPUs under 24 GB. Set OLLAMA_CONTEXT_LENGTH when starting the server, or /set parameter num_ctx inside the chat.
  • /set parameter num_ctx 8192

Related pages

Frequently asked questions

Can I run Qwen3 30B-A3B (2507) on the GeForce RTX 5090?

Q4_K_M fits in VRAM with 11.5 GB to spare: Runs great, est. 235.0 tok/s (188.0–282.0, calibrated estimate ±20%). Q4_K_M needs 19.9 GB at 8k context; this setup has 31.4 GB of usable memory and 28.0 GB of free system RAM.

How much context can Qwen3 30B-A3B (2507) use on the GeForce RTX 5090?

At Q4_K_M the verdict stays Runs great up to 64k context with an f16 KV cache (6.4 GB of KV) and up to 128k with a q8_0 KV cache (6.9 GB). The model supports up to 256k.

Which quant should I use, and how fast is it?

Q4_K_M (18.56 GB) is the recommended balance: Runs great, est. 235.0 tok/s (188.0–282.0, calibrated estimate ±20%). It is also the largest tracked file that gets this verdict.

Speeds are estimates from memory bandwidth, calibrated against public benchmarks, and each one comes with an error band and a confidence label. Real results vary with drivers, backend, context length and thermals.