Can RTX 3090 run Qwen3.6 27B?
Yes. Measured 42.7 tok/s generation (llama.cpp 0.7.0-dev, Q4_K_M; source).
Every measured result for Qwen3.6 27B on RTX 3090
| Build | Quant / setup | Engine | tok/s | |
|---|---|---|---|---|
| RTX 3090 (llm-speed signed runs, Oct 2026): gpt-oss 20B 207 tok/s, Qwen3.6 27B 43 tok/s | Q4_K_M | llama.cpp 0.7.0-dev | 42.7 | source |
| RTX 3090: 2026 models (llamabench community) | llama-bench tg128 (quant per result) | llama.cpp | 40.2 | source |
Memory: RTX 3090 has 24 GB. Qwen3.6 27B weights are about 16.7 GB at Q4_K_M (27.32B params × 4.89 bits ÷ 8, before context/KV cache). Arithmetic, not a measurement; long contexts need more. Check other quants in the VRAM calculator.
Speeds are decode (generation) tokens/sec as reported by each source; quant, engine and context differ, so small gaps are noise.
All measured speeds for Qwen3.6 27B: best home rig for Qwen3.6 27B.
Qwen3.6 27B on other hardware: DGX Spark (GB10) · RTX 4090 · RTX 5090 · Radeon AI Pro R9700 · Strix Halo (Radeon 8060S)
Other models measured on RTX 3090: DeepSeek R1 671B · Flint 40B-A2.5B · GLM 4.7 Flash · Qwen3.6 35B-A3B · Qwen3.8 27B · gpt-oss 20B
Not enough VRAM? Rent before you buy. Run Qwen3.6 27B on a bigger GPU on a rented 24–96 GB cloud GPU by the hour and see the real speed before spending €1,000+. Try it on Runpod
Referral link: if you sign up with Google and add $10, Runpod gives you and RigBase bonus credit (terms). It never affects our data.
Referral link: if you sign up with Google and add $10, Runpod gives you and RigBase bonus credit (terms). It never affects our data.