LLM Roofline
NVIDIA · PCIE · Backend: llama.cpp

NVIDIA RTX 3060 (12 GB) — LLM Sizing & Speed Matrix

Most popular entry 12 GB GPU for 8B Q6/Q8 and 14B IQ4_XS tight context.

Reference Pricing
$289 Street
Usable Memory
11 GiB
Memory Bandwidth
360 GB/s (78% eff.)
FP16 Compute
25.5 TFLOPS
Roofline Ridge
54.5 FLOP/B
Power (Load / Idle)
170W / 20W

Buy or Rent NVIDIA RTX 3060 (12 GB)

Vendor & Cloud Links
RTX 3060 12 GB on Amazon
Amazon · Direct Hardware
↗
NVIDIA RTX Cards on Newegg
Newegg · Direct Hardware
↗

All 32 Open-Weight Models on NVIDIA RTX 3060 (12 GB) (Q4_K_M @ 32k)

Model Status Weights Total @ 32k Decode TTFT (4k) Max Safe Ctx
Qwen2.5 7B Instruct COMFORTABLE 4.4 GiB 7.4 GiB 42.7 tok/s 4.32s 32k
Gemma 3 4B IT COMFORTABLE 2.5 GiB 4.4 GiB 78.4 tok/s 2.49s 128k
Gemma 3 12B IT TIGHT 7.0 GiB 10.6 GiB 28.0 tok/s 7.00s 38k
Qwen3 8B TIGHT 4.7 GiB 10.5 GiB 28.4 tok/s 4.75s 35k
Llama 3.1 8B Instruct TIGHT 4.6 GiB 9.9 GiB 30.3 tok/s 4.62s 41k
DeepSeek-R1 671B OOM 386.4 GiB 390.3 GiB OOM — OOM
DeepSeek-V3 671B OOM 386.4 GiB 390.3 GiB OOM — OOM
Llama 3.1 405B Instruct OOM 233.8 GiB 252.5 GiB OOM — OOM
Llama 4 Maverick 400B-A17B OOM 231.5 GiB 239.0 GiB OOM — OOM
Qwen3 235B-A22B OOM 135.4 GiB 142.6 GiB OOM — OOM
Mixtral 8x22B Instruct OOM 81.2 GiB 89.8 GiB OOM — OOM
Command A 111B OOM 63.9 GiB 69.1 GiB OOM — OOM
Llama 4 Scout 109B-A17B OOM 62.8 GiB 70.2 GiB OOM — OOM
Qwen2.5 72B Instruct OOM 41.9 GiB 53.7 GiB OOM — OOM
DeepSeek-R1 Distill Llama 70B OOM 40.7 GiB 52.5 GiB OOM — OOM
Llama 3.3 70B Instruct OOM 40.7 GiB 52.5 GiB OOM — OOM
Llama 3.1 70B Instruct OOM 40.7 GiB 52.5 GiB OOM — OOM
Llama-3.3 Nemotron Super 49B OOM 28.7 GiB 36.7 GiB OOM — OOM
Mixtral 8x7B Instruct OOM 26.9 GiB 32.2 GiB OOM — OOM
Qwen3 32B OOM 18.9 GiB 28.3 GiB OOM — OOM
Qwen2.5 Coder 32B OOM 18.9 GiB 28.3 GiB OOM — OOM
DeepSeek-R1 Distill Qwen 32B OOM 18.9 GiB 28.3 GiB OOM — OOM
QwQ 32B Reasoning OOM 18.9 GiB 28.3 GiB OOM — OOM
GLM-4 32B 0414 OOM 18.6 GiB 22.1 GiB OOM — OOM
Qwen3 30B-A3B OOM 17.6 GiB 21.6 GiB OOM — OOM
Gemma 3 27B IT OOM 15.8 GiB 20.2 GiB OOM — OOM
Mistral Small 3.1 24B OOM 13.8 GiB 20.3 GiB OOM — OOM
Codestral 22B OOM 12.8 GiB 21.4 GiB OOM — OOM
Qwen3 14B OOM 8.5 GiB 15.0 GiB OOM — 7k
Qwen2.5 14B Instruct OOM 8.5 GiB 16.0 GiB OOM — 6k
DeepSeek-R1 Distill Qwen 14B OOM 8.5 GiB 16.0 GiB OOM — 6k
Phi-4 14B OOM 8.5 GiB 13.0 GiB OOM — 6k

Interactive NVIDIA RTX 3060 (12 GB) Simulator

7.6B Dense · GQA
360 GB/s · 25.5 TFLOPS
3. Weight Quantization 4.85 bpw
4. KV Cache
32,768 tok
Compute Roofline
COMFORTABLE FIT Memory-Bandwidth Bound
Max Safe Ctx: 32k
Memory Allocation Footprint 7.37 / 11.00 GiB usable
Weights: 4.38 GiB KV Cache: 1.75 GiB Overhead: 1.24 GiB Headroom: 3.63 GiB
Decode Speed
42.7
tok/s (batch=1)
Prefill Speed
955
prompt tok/s
TTFT Latency
4.32s
for 4,096 tok
Active Params
7.6B
of 7.6B total
llama.cpp Server Command 
llama-server -hf Qwen/Qwen2.5-7B-Instruct:Q4_K_M -ngl 99 -fa on -c 32768 --port 8080
Context KV Cache Total VRAM Decode Status
4k 0.22 GiB 5.59 GiB 56.9 tok/s COMFORTABLE
8k 0.44 GiB 6.06 GiB 54.3 tok/s COMFORTABLE
16k 0.88 GiB 6.49 GiB 49.8 tok/s COMFORTABLE
32k 1.75 GiB 7.37 GiB 42.7 tok/s COMFORTABLE

Buy-vs-Cloud-vs-API Monthly TCO Break-Even

USE SERVERLESS API
Own Hardware/mo
$10.87
112.2h active + power
Cloud GPU/mo
$114.81
@$0.89/hr spot/pod
Serverless API/mo
$3.80
API cheaper

Recommended Hardware & Cloud Deployment

Contextual to fit status

Hardware FAQ — NVIDIA RTX 3060 (12 GB)

How much usable VRAM / GPU memory does NVIDIA RTX 3060 (12 GB) actually provide for LLMs?

NVIDIA RTX 3060 (12 GB) provides 11 GiB usable VRAM after subtracting 1.0 GiB driver/display overhead.

What is the largest Q4_K_M model NVIDIA RTX 3060 (12 GB) can run at 32k context?

With Q4_K_M weights and FP16 KV cache at 32k context, the largest fitting model in our verified catalog on NVIDIA RTX 3060 (12 GB) is Qwen2.5 7B Instruct (7.6B total).