GeForce RTX 4060
8 GB GDDR6 · 272 GB/s · 115 W
RTX 4060 8 GB · 64 GB RAM → Qwen3 8B
GPU · NVIDIA
RTX 4060 for local LLMs: which models run?
With 8 GB of VRAM and 272 GB/s of bandwidth, the RTX 4060 runs models up to about Qwen3 8B entirely on the GPU.
- VRAM8 GB GDDR6
- Bandwidth272 GB/s
- Board power115 W
- ArchitectureAda Lovelace
Specifications
- VRAM
- 8 GB
- Memory type
- GDDR6
- Memory bus
- 128-bit
- Bandwidth
- 272 GB/s
- Board power
- 115 W
- Backends
- CUDA, Vulkan
- Interface
- PCIe 4.0 x8
- Released
- June 2023
- Segment
- Consumer
- FP16 compute
- Not available
Official specData updated 24 days agoSource: NVIDIA specifications
What the RTX 4060 runs
| Gemma 3 4B ITQ8_0 · 8K~5.4 GB · 45–60 tok/s estimate | Excellent |
|---|---|
| Qwen3 4BQ8_0 · 8K~6.2 GB · 39–53 tok/s estimate | Excellent |
| Phi-4 Mini Instruct 3.8BQ8_0 · 8K~6 GB · 42–56 tok/s estimate | Excellent |
| Llama 3.2 3B InstructQ8_0 · 8K~5.1 GB · 50–66 tok/s estimate | Excellent |
| Qwen3 1.7BQ8_0 · 8K~3.7 GB · 79–109 tok/s estimate | Excellent |
| Llama 3.2 1B InstructQ8_0 · 8K~2.5 GB · 127–174 tok/s estimate | Excellent |
| Qwen3 0.6BQ8_0 · 8K~2.5 GB · 144–213 tok/s estimate | Excellent |
| Qwen3 8BQ4_K_M · 8K~6.9 GB · 35–46 tok/s estimate | Compatible |
| Llama 3.1 8B InstructQ4_K_M · 8K~6.7 GB · 36–48 tok/s estimate | Compatible |
| DeepSeek R1 Distill Llama 8BQ4_K_M · 8K~6.7 GB · 36–48 tok/s estimate | Compatible |
| Qwen2.5 7B InstructQ5_K_M · 8K~6.6 GB · 35–46 tok/s estimate | Compatible |
| Mistral 7B Instruct v0.3Q5_K_M · 8K~6.7 GB · 35–46 tok/s estimate | Compatible |
| Mistral Small 3.2 24B InstructQ3_K_M · 8K~13.3 GB · 3.5–6 tok/s estimate | CPU/GPU offload |
| gpt-oss-20bMXFP4 · 8K~13 GB · 18–33 tok/s estimate | CPU/GPU offload |
| Qwen3 14BQ3_K_M · 8K~9.3 GB · 8.5–16 tok/s estimate | CPU/GPU offload |
| DeepSeek R1 Distill Qwen 14BQ3_K_M · 8K~9.6 GB · 8–15 tok/s estimate | CPU/GPU offload |
| Qwen2.5 14B InstructQ3_K_M · 8K~9.6 GB · 8–15 tok/s estimate | CPU/GPU offload |
| Phi-4 14BQ3_K_M · 8K~9.6 GB · 8–15 tok/s estimate | CPU/GPU offload |
| Mistral Nemo 12B InstructQ3_K_M · 8K~8.1 GB · 17–29 tok/s estimate | CPU/GPU offload |
| Gemma 3 12B ITQ3_K_M · 8K~7.8 GB · 21–36 tok/s estimate | CPU/GPU offload |
| Gemma 2 9B ITQ3_K_M · 8K~7.8 GB · 24–41 tok/s estimate | CPU/GPU offload |
| Llama 4 Scout 17B-16E InstructQ3_K_M · 8K~55.7 GB · 2.5–4.5 tok/s estimate | Possible but slow |
| Mixtral 8x7B InstructQ3_K_M · 8K~21.6 GB · 4.5–9 tok/s estimate | Possible but slow |
| Qwen3 32BQ3_K_M · 8K~18.6 GB · 1.8–3.5 tok/s estimate | Possible but slow |
| DeepSeek R1 Distill Qwen 32BQ3_K_M · 8K~18.5 GB · 1.8–3.5 tok/s estimate | Possible but slow |
| Qwen2.5 32B InstructQ3_K_M · 8K~18.5 GB · 1.8–3.5 tok/s estimate | Possible but slow |
| Qwen2.5 Coder 32B InstructQ3_K_M · 8K~18.5 GB · 1.8–3.5 tok/s estimate | Possible but slow |
| QwQ 32BQ3_K_M · 8K~18.5 GB · 1.8–3.5 tok/s estimate | Possible but slow |
| Qwen3 30B-A3BQ3_K_M · 8K~15.4 GB · 17–32 tok/s estimate | Possible but slow |
| Qwen3 Coder 30B-A3B InstructQ3_K_M · 8K~16 GB · 16–30 tok/s estimate | Possible but slow |
| Gemma 3 27B ITQ3_K_M · 8K~15.3 GB · 2.5–5 tok/s estimate | Possible but slow |
| Gemma 2 27B ITQ3_K_M · 8K~16.5 GB · 2.5–4.5 tok/s estimate | Possible but slow |
| Qwen2.5 72B InstructQ3_K_M · 8K~38.2 GB · 0.7–1.3 tok/s estimate | Not recommended |
| Llama 3.3 70B InstructQ3_K_M · 8K~37 GB · 0.7–1.3 tok/s estimate | Not recommended |
| DeepSeek R1 Distill Llama 70BQ3_K_M · 8K~37 GB · 0.7–1.3 tok/s estimate | Not recommended |
| DeepSeek R1 (671B)Q8_0 · 8K~699 GB · No data | Incompatible |
| Qwen3 235B-A22BQ8_0 · 8K~247 GB · No data | Incompatible |
| gpt-oss-120bMXFP4 · 8K~63.2 GB · No data | Incompatible |
For each model, the highest-precision quantization that runs entirely in fast memory; otherwise the one with the best result. Speeds are ranges; "estimate" unless a benchmark exists for this exact hardware.
EstimateData updated 24 days agoSource: Run LLM Locally engine
Limits in plain words
8 GB of VRAM holds the weights and KV cache of models up to about Qwen3 8B at Q4_K_M. Bigger models spill into system RAM and slow down sharply: generation speed is then limited by RAM bandwidth, not by the GPU.
Where to buy a RTX 4060
Prices and links follow the market you choose.
No current offer imported for this market: the link opens a marketplace search.
Frequently asked questions
Which LLMs can the RTX 4060 run?
Entirely on the GPU: Qwen3 8B (Q4_K_M), Llama 3.1 8B Instruct (Q4_K_M), DeepSeek R1 Distill Llama 8B (Q4_K_M), Qwen2.5 7B Instruct (Q5_K_M), Mistral 7B Instruct v0.3 (Q5_K_M), Gemma 3 4B IT (Q8_0), Qwen3 4B (Q8_0), Phi-4 Mini Instruct 3.8B (Q8_0). Larger models need CPU/GPU offload.
Can the RTX 4060 run a 70B model?
Not recommended: a 70B model at Q4_K_M needs about 45.1 GB; the RTX 4060 has 8 GB.
How is the speed estimated?
Token generation reads every active weight once per token, so it is limited by memory bandwidth (272 GB/s here). We apply an efficiency range per backend and label the result as an estimate until real benchmarks exist.