Can I run Qwen3 32B on the GeForce RTX 4060?
Won't runQ4_K_M fits in memory (22.0 GB) but is too slow to use: est. 2.5 tok/s (2.0–3.0, calibrated estimate ±20%). Q2_K (heavy quality loss) works instead: Runs slowly, est. 4.4 tok/s (3.5–5.2, calibrated estimate ±20%).
8 GB VRAM · 272.0 GB/s · FP16 15.1 TFLOPS. Assumes 32 GB of DDR5-5600 system RAM, Windows with this GPU driving the display, 8k context and an f16 KV cache. The verdict and quant file above stay the same anywhere from 16 to 128 GB of system RAM.
Every quant of Qwen3 32B on the GeForce RTX 4060
| Quant | File size | Verdict | Speed | Memory | Runs as | Notes |
|---|---|---|---|---|---|---|
| Q2_K | 12.34 GB | Runs slowly | est. 4.4 tok/s3.5–5.2calibrated estimate ±20% | 8.0 / 8.0 GB + 7.7 GB RAM | Partial offload | — |
| IQ4_XS | 17.69 GB | Won't run | est. 2.8 tok/s2.3–3.4calibrated estimate ±20% | 8.0 / 8.0 GB + 13.0 GB RAM | Partial offload | Fits, but too slow to use |
| Q4_K_Mbaseline | 19.76 GB | Won't run | est. 2.5 tok/s2.0–3.0calibrated estimate ±20% | 8.0 / 8.0 GB + 15.1 GB RAM | Partial offload | Fits, but too slow to use |
| Q8_0 | 34.82 GB | Won't run | — | Needs about 36.7 GB; 7.4 GB of VRAM and 28.0 GB of RAM are free | — | — |
Models with a thinking mode spend extra tokens thinking, so their speed thresholds are 1.5× stricter (30 / 12 / 3 tok/s). With thinking turned off, a model may do one grade better.
Where the memory goes at Q2_K
- Weights
- 4.7 GB
- KV cache
- 2.1 GB
- Compute buffer
- 0.6 GB
- OS reserve
- 0.6 GB
- Weights in system RAM
- 7.7 GB
Context length vs. KV cache
| Context | KV f16 | KV q8_0 | KV q4_0 |
|---|---|---|---|
| 4k | 13.7 GBRuns slowly est. 5.2 tok/s4.2–6.2calibrated estimate ±20% | 13.2 GBRuns slowly est. 5.7 tok/s4.5–6.8calibrated estimate ±20% | 12.9 GBRuns slowly est. 6.0 tok/s4.8–7.2calibrated estimate ±20% |
| 8k | 14.8 GBRuns slowly est. 4.4 tok/s3.5–5.2calibrated estimate ±20% | 13.8 GBRuns slowly est. 5.1 tok/s4.1–6.1calibrated estimate ±20% | 13.3 GBRuns slowly est. 5.6 tok/s4.5–6.7calibrated estimate ±20% |
| 16k | 17.0 GBRuns slowly est. 3.2 tok/s2.5–3.8calibrated estimate ±20% | 15.0 GBRuns slowly est. 4.2 tok/s3.4–5.1calibrated estimate ±20% | 14.0 GBRuns slowly est. 5.0 tok/s4.0–6.0calibrated estimate ±20% |
| 32k | 20.9 GBWon't run est. 2.1 tok/s1.7–2.6calibrated estimate ±20% | 17.5 GBRuns slowly est. 3.0 tok/s2.4–3.6calibrated estimate ±20% | 15.3 GBRuns slowly est. 4.1 tok/s3.3–4.9calibrated estimate ±20% |
S = Runs great · A = Runs well · B = Runs slowly · F = Won't run
At Q2_K the verdict stays Runs slowly up to 16k context with an f16 KV cache, and up to 32k with q8_0.
Estimated speed
- Generation
- est. 4.4 tok/s (3.5–5.2, calibrated estimate ±20%)
- Prompt processing
- Prompt-processing estimates only apply when the whole model runs on the GPU.
Measured on this exact combination
No public measurement for this exact combination yet; the numbers above are estimates.
If this is not enough
A smaller model that runs well on the GeForce RTX 4060
Runs wellNemotron 3.5 Lightning 30B-A3B: Runs well, est. 32.6 tok/s (22.8–42.3, theoretical estimate ±30%)
A lower quant of Qwen3 32B
Runs slowlyQ2_K (12.34 GB, heavy quality loss): Runs slowly, est. 4.4 tok/s (3.5–5.2, calibrated estimate ±20%)
Cheapest GPU that runs Qwen3 32B great
Runs greatGeForce RTX 4090 — $2,755 street (as of 2026-08-09): 42.5 tok/s measured at 512 context (1 run; 8k estimate 32.2 tok/s, 28.3–36.1, calibrated estimate ±12%).
How to run it
Commands for the Q2_K file. Both tools download from Hugging Face on first run.
File: Qwen_Qwen3-32B-Q2_K.gguf (12.34 GB) from bartowski/Qwen_Qwen3-32B-GGUF.
llama-server -hf bartowski/Qwen_Qwen3-32B-GGUF:Q2_K -c 8192 -ngl 24 -fa on- -ngl 24 puts 24 of 64 layers on the GPU; if VRAM overflows, lower it by 1–2.
- -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 1.15 GB at 8k context.
PowerShell (quit the Ollama tray app first):
$env:OLLAMA_CONTEXT_LENGTH="8192"; ollama serve
ollama run hf.co/bartowski/Qwen_Qwen3-32B-GGUF:Q2_K- 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.
- Ollama splits layers between GPU and CPU automatically.
/set parameter num_ctx 8192
Related pages
Guide: How much VRAM do local LLMs need? GPU requirements for 8–32 GB (2026)
Other models on the GeForce RTX 4060
Qwen3 32B on other hardware
Frequently asked questions
Can I run Qwen3 32B on the GeForce RTX 4060?
Q4_K_M fits in memory (22.0 GB) but is too slow to use: est. 2.5 tok/s (2.0–3.0, calibrated estimate ±20%). Q2_K (heavy quality loss) works instead: Runs slowly, est. 4.4 tok/s (3.5–5.2, calibrated estimate ±20%). Q2_K needs 14.8 GB at 8k context; this setup has 7.4 GB of usable memory and 28.0 GB of free system RAM.
How much context can Qwen3 32B use on the GeForce RTX 4060?
At Q2_K the verdict stays Runs slowly up to 16k context with an f16 KV cache (4.3 GB of KV) and up to 32k with a q8_0 KV cache (4.6 GB). The model supports up to 32k.
Which quant should I use, and how fast is it?
Q2_K (heavy quality loss) (12.34 GB) is the recommended balance: Runs slowly, est. 4.4 tok/s (3.5–5.2, 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.