Custom build engine

Describe your build — any GPUs, CPU, RAM, OS, runtime, use case. We'll compute effective VRAM honestly, recommend a runtime, and tell you which models fit comfortably, which are borderline, and which aren't practical.

Total VRAM ≠ pooled VRAM. We never sum VRAM unless the silicon truly pools (Apple unified memory). We always explain why effective is lower than total.

Calculations follow the RunLocalAI Will-It-Run Framework: effective VRAM, model working set, runtime constraints, fit tiers, and measured-vs-estimated evidence labels.

Describe your build

Add GPUs, set CPU/RAM/OS, optionally pick a runtime + use case. URL updates as you change fields — share a build by copying the URL.

GPUs in your build

CPU class

System RAM (GB)

Use case

Your skill level

Runtime preference

Build summary

Total VRAM

16 GB

Effective VRAM

~14 GB

range 13-14 GB

Topology

single gpu

none

Setup difficulty

beginner

speed penalty ~0%

Why effective VRAM is lower than total

Single NVIDIA GeForce RTX 3080 16GB (Mobile) — 16 GB VRAM minus ~1.8 GB runtime/driver overhead = ~14 GB usable for weights + KV cache + activations. The remaining uncertainty band covers OS display use and background CUDA allocations.

Measured evidence on this hardware

Publicly inspectable measured rows for the selected hardware slug(s). Exact measured rows calibrate the fit table instead of leaving it as pure VRAM estimation.

Speed rows

12 measured

Model	Evidence	Quant	Tok/s	Provenance
Turkcell LLM 7B v1 NVIDIA GeForce RTX 3080 16GB (Mobile)	4K ctx ollama version is 0.24.0 Microsoft Windows [Version 10.0.26200.8457] Driver 571.96	Q4_K_M	85.77	Measured hererun logindependent reproduction pending
RefinedNeuro RN TR R2 NVIDIA GeForce RTX 3080 16GB (Mobile)	4K ctx ollama version is 0.24.0 Microsoft Windows [Version 10.0.26200.8457] Driver 571.96	Q4_K_M	79.27	Measured hererun logindependent reproduction pending
RefinedNeuro RN TR R1 NVIDIA GeForce RTX 3080 16GB (Mobile)	4K ctx ollama version is 0.24.0 Microsoft Windows [Version 10.0.26200.8457] Driver 571.96	Q4_K_M	79.89	Measured hererun logindependent reproduction pending
Qwen 3 4B NVIDIA GeForce RTX 3080 16GB (Mobile)	4K ctx ollama version is 0.24.0 Microsoft Windows [Version 10.0.26200.8457] Driver 571.96	Q4_K_M	103.7	Measured hererun logindependent reproduction pending
Qwen 3 14B NVIDIA GeForce RTX 3080 16GB (Mobile)	4K ctx ollama version is 0.24.0 Microsoft Windows [Version 10.0.26200.8457] Driver 571.96	Q4_K_M	38.26	Measured hererun logindependent reproduction pending
Qwen 2.5 7B Instruct NVIDIA GeForce RTX 3080 16GB (Mobile)	4K ctx ollama version is 0.24.0 Microsoft Windows [Version 10.0.26200.8457] Driver 571.96	Q4_K_M	80.42	Measured hererun logindependent reproduction pending
Phi-4 Reasoning 14B NVIDIA GeForce RTX 3080 16GB (Mobile)	4K ctx ollama version is 0.24.0 Microsoft Windows [Version 10.0.26200.8457] Driver 571.96	Q4_K_M	40.44	Measured hererun logindependent reproduction pending
Phi-3.5 Mini Instruct NVIDIA GeForce RTX 3080 16GB (Mobile)	4K ctx ollama version is 0.24.0 Microsoft Windows [Version 10.0.26200.8457] Driver 571.96	Q4_K_M	155.4	Measured hererun logindependent reproduction pending
Mistral Nemo 12B Instruct NVIDIA GeForce RTX 3080 16GB (Mobile)	4K ctx ollama version is 0.24.0 Microsoft Windows [Version 10.0.26200.8457] Driver 571.96	Q4_K_M	65.72	Measured hererun logindependent reproduction pending
Mistral 7B Instruct v0.3 NVIDIA GeForce RTX 3080 16GB (Mobile)	4K ctx ollama version is 0.24.0 Microsoft Windows [Version 10.0.26200.8457] Driver 571.96	Q4_K_M	89.64	Measured hererun logindependent reproduction pending
Llama 3.2 11B Vision Instruct NVIDIA GeForce RTX 3080 16GB (Mobile)	4K ctx ollama version is 0.24.0 Microsoft Windows [Version 10.0.26200.8457] Driver 571.96	Q4_K_M	67.00	Measured hererun logindependent reproduction pending
Malhajar Mistral 7B Turkish NVIDIA GeForce RTX 3080 16GB (Mobile)	4K ctx ollama version is 0.24.0 Microsoft Windows [Version 10.0.26200.8457] Driver 571.96	Q4_K_M	87.28	Measured hererun logindependent reproduction pending

Quality rows

12 public

Model	Benchmark	Setup	Score	Log
Ministral 3 14B	HumanEval+	Q4_K_M / ollama-0.32	76.8	gist
Qwen3.5 9B	HumanEval+	Q4_K_M / ollama-0.32	42.7	gist
Ornith 1.0 9B	HumanEval+	Q4_K_M / ollama-0.32	73.2	gist
Qwen 3 8B	HumanEval+	Q4_K_M / ollama-0.24	2.4	gist
Llama 3.1 8B Instruct	MBPP+	Q4_K_M / ollama-0.24	39.2	gist
Phi-4 14B	MBPP+	Q4_K_M / ollama-0.24	60.3	gist
Qwen 2.5 Coder 7B Instruct	MBPP+	Q4_K_M / ollama-0.24	66.9	gist
Llama 3.1 8B Instruct	HumanEval+	Q4_K_M / ollama-0.24	56.1	gist
Phi-4 14B	HumanEval+	Q4_K_M / ollama-0.24	78.7	gist
Qwen 2.5 Coder 7B Instruct	HumanEval+	Q4_K_M / ollama-0.24	81.1	gist
Llama 3.2 3B Instruct	TurkishMMLU (Generative)	Q4_K_M / ollama-0.24	11.4	gist
Turkish Llama 8B Instruct v0.1	TurkishMMLU (Generative)	Q4_K_M / ollama-0.24	11.0	gist

Recommended runtime

Best engine for this topology + skill level + use case.

ExLlamaV2

primary

involved

Highest single-stream throughput on consumer NVIDIA. EXL2 mixed-bit quants are the leading consumer-tier inference format.

llama.cpp

alternative

moderate

Cross-format flexibility — GGUF works everywhere; the engine that powers Ollama and LM Studio.

WORKLOAD PROFILE

FITS

OLMo 2 13B @ Q4_K_M, 4K context on NVIDIA GeForce RTX 3080 16GB (Mobile)

0 GB16 GBVRAM ceiling

Weights7.8 GB

KV cache3.3 GB

Activations0.4 GB

Runtime1.8 GB

Headroom2.8 GB

ESTIMATED DECODE RATE

53 tok/s

Bandwidth-derived estimate · efficiency 0.80. Real-world rates land within ±20% on well-tuned runtimes.

Models that fit your build

345 models considered. Categorized by headroom at the recommended quant + a sensible context for your use case.

Comfortable

24 models · ≥15% headroom

Model	Params	Quant	VRAM est.	Context	Evidence	Note
OLMo 2 13B	13B	Q4_K_M	11.4 GB	4,096	No measured row yet	Fits cleanly at Q4_K_M + 4,096 ctx with 18% headroom.
OpenThaiGPT 1.0.0 Beta 13B Chat	13B	Q4_K_M	10.8 GB	4,096	No measured row yet	Fits cleanly at Q4_K_M + 4,096 ctx with 23% headroom.
mGPT 13B	13B	Q4_K_M	9.2 GB	2,048	No measured row yet	Fits cleanly at Q4_K_M + 2,048 ctx with 34% headroom.
Stable LM 2 12B	12B	Q4_K_M	10.6 GB	4,096	No measured row yet	Fits cleanly at Q4_K_M + 4,096 ctx with 25% headroom.
FLUX.1 [dev]	12B	Q4_K_M	6.9 GB	0	No measured row yet	Comfortable fit with 51% headroom — room to extend context or run alongside other workloads.
FLUX.1 [schnell]	12B	Q4_K_M	6.9 GB	0	No measured row yet	Comfortable fit with 51% headroom — room to extend context or run alongside other workloads.
Merlyn Education Safety 12B AWQ	12B	Q4_K_M	8.4 GB	2,048	No measured row yet	Fits cleanly at Q4_K_M + 2,048 ctx with 40% headroom.
Trendyol LLM Asure 12B	12B	GGUF_UNKNOWN	10.9 GB	8,192	No measured row yet	Fits cleanly at GGUF_UNKNOWN + 8,192 ctx with 22% headroom.
Bielik 11B v2.3 Instruct	11B	Q4_K_M	9.2 GB	4,096	No measured row yet	Fits cleanly at Q4_K_M + 4,096 ctx with 35% headroom.
Bielik 11B v2.3 Instruct	11B	Q4_K_M	9.2 GB	4,096	No measured row yet	Fits cleanly at Q4_K_M + 4,096 ctx with 35% headroom.
Bielik-11B v3.0 Instruct FP8 Dynamic	11B	Q4_K_M	9.2 GB	4,096	No measured row yet	Fits cleanly at Q4_K_M + 4,096 ctx with 35% headroom.
SOLAR 10.7B v1.0	11B	Q4_K_M	8.9 GB	4,096	No measured row yet	Fits cleanly at Q4_K_M + 4,096 ctx with 37% headroom.
Falcon 3 10B	10B	Q4_K_M	11.3 GB	8,192	No measured row yet	Fits cleanly at Q4_K_M + 8,192 ctx with 19% headroom.
YTU Turkish Gemma 9B v0.1	9B	Q4_K_M	10.7 GB	8,192	No measured row yet	Fits cleanly at Q4_K_M + 8,192 ctx with 24% headroom.
NVIDIA Nemotron Nano 9B v2 Japanese	9B	Q4_K_M	9.8 GB	8,192	No measured row yet	Fits cleanly at Q4_K_M + 8,192 ctx with 30% headroom.
Nemotron 3 Nano 9B	9B	Q4_K_M	10.1 GB	8,192	No measured row yet	Fits cleanly at Q4_K_M + 8,192 ctx with 28% headroom.
Gemma 2 9B Instruct	9B	Q4_K_M	10.6 GB	8,192	No measured row yet	Fits cleanly at Q4_K_M + 8,192 ctx with 24% headroom.
Turkish Gemma 9B T1	9B	Q4_K_M	9.8 GB	8,192	No measured row yet	Fits cleanly at Q4_K_M + 8,192 ctx with 30% headroom.
Yi Coder 9B	9B	Q4_K_M	10.2 GB	8,192	No measured row yet	Fits cleanly at Q4_K_M + 8,192 ctx with 27% headroom.
GLM-4 9B	9B	Q4_K_M	10.3 GB	8,192	No measured row yet	Fits cleanly at Q4_K_M + 8,192 ctx with 27% headroom.
Qwen3.5 9B	9B	Q4_K_M	11.4 GB	8,192	No measured row yet	Fits cleanly at Q4_K_M + 8,192 ctx with 18% headroom.
Ornith 1.0 9B	9B	Q4_K_M	10.4 GB	8,192	No measured row yet	Fits cleanly at Q4_K_M + 8,192 ctx with 26% headroom.
Qwen3.5 9B Thai Law Base	9B	Q4_K_M	7.4 GB	4,096	No measured row yet	Comfortable fit with 47% headroom — room to extend context or run alongside other workloads.
Granite 4.1 8B Instruct	9B	Q5_K_M	10.7 GB	8,192	No measured row yet	Fits cleanly at Q5_K_M + 8,192 ctx with 23% headroom.

Borderline

14 models · tight, may need quant downgrade

Model	Params	Quant	VRAM est.	Context	Evidence	Note
DeepSeek MoE 16B Base	16B	Q4_K_M	14 GB	4,096	No measured row yet	Tight fit at Q4_K_M — only 0% headroom. KV cache for longer context will OOM. Cap context tighter or drop one quant level.
Phi-4 Reasoning 14B	14B	Q4_K_M	15.8 GB	8,192	40.44 tok/sMeasured hereNVIDIA GeForce RTX 3080 16GB (Mobile) / 4K ctx	Measured on this hardware: 40.44 tok/s at Q4_K_M (4,096 measured ctx). The larger target context still needs validation before calling it comfortable.
Qwen 3 14B	14B	Q4_K_M	15.8 GB	8,192	38.26 tok/sMeasured hereNVIDIA GeForce RTX 3080 16GB (Mobile) / 4K ctx	Measured on this hardware: 38.26 tok/s at Q4_K_M (4,096 measured ctx). The larger target context still needs validation before calling it comfortable.
Pixtral 12B	12B	Q4_K_M	13.4 GB	8,192	No measured row yet	Tight fit at Q4_K_M — only 5% headroom. KV cache for longer context will OOM. Cap context tighter or drop one quant level.
Mistral Nemo 12B Instruct	12B	Q4_K_M	13.9 GB	8,192	65.72 tok/sMeasured hereNVIDIA GeForce RTX 3080 16GB (Mobile) / 4K ctx	Measured on this hardware: 65.72 tok/s at Q4_K_M (4,096 measured ctx). Tight fit at Q4_K_M — only 1% headroom. KV cache for longer context will OOM. Cap context tighter or drop one quant level.
Gemma 3 12B	12B	Q4_K_M	13.7 GB	8,192	No measured row yet	Tight fit at Q4_K_M — only 2% headroom. KV cache for longer context will OOM. Cap context tighter or drop one quant level.
Gemma 4 12B	12B	Q4_K_M	14 GB	8,192	No measured row yet	Tight fit at Q4_K_M — only 0% headroom. KV cache for longer context will OOM. Cap context tighter or drop one quant level.
Bielik 11B v2.2 Instruct GGUF	11B	Q4_K_M	11.9 GB	8,192	No measured row yet	Tight fit at Q4_K_M — only 15% headroom. KV cache for longer context will OOM. Cap context tighter or drop one quant level.
Llama 3.2 11B Vision Instruct	11B	Q4_K_M	13.8 GB	8,192	67.00 tok/sMeasured hereNVIDIA GeForce RTX 3080 16GB (Mobile) / 4K ctx	Measured on this hardware: 67.00 tok/s at Q4_K_M (4,096 measured ctx). Tight fit at Q4_K_M — only 1% headroom. KV cache for longer context will OOM. Cap context tighter or drop one quant level.
Llama 3.2 11B Vision	11B	Q4_K_M	12.3 GB	8,192	No measured row yet	Tight fit at Q4_K_M — only 12% headroom. KV cache for longer context will OOM. Cap context tighter or drop one quant level.
Bielik 11B v3.0 Instruct GGUF	11B	Q4_K_M	11.9 GB	8,192	No measured row yet	Tight fit at Q4_K_M — only 15% headroom. KV cache for longer context will OOM. Cap context tighter or drop one quant level.
Hermes 3 Llama 3.1 8B	8B	Q8_0	12.9 GB	8,192	No measured row yet	Tight fit at Q8_0 — only 8% headroom. KV cache for longer context will OOM. Cap context tighter or drop one quant level.
Qwen 3 8B	8B	Q8_0	12.6 GB	8,192	No measured row yet	Tight fit at Q8_0 — only 10% headroom. KV cache for longer context will OOM. Cap context tighter or drop one quant level.
DeepSeek R1 Distill Qwen 7B	7B	Q8_0	12 GB	8,192	No measured row yet	Tight fit at Q8_0 — only 14% headroom. KV cache for longer context will OOM. Cap context tighter or drop one quant level.

Not practical

16 models · oversize for this build

Model	Params	Quant	VRAM est.	Context	Evidence	Note
NV-Embed v2	8B	FP16	19.7 GB	8,192	No measured row yet	~19.7 GB needed at FP16 + 8,192 ctx — overshoots effective VRAM by 41%. Drop quant or move to a larger build.
Qwen 3 Embedding 8B	8B	FP16	20.8 GB	8,192	No measured row yet	~20.8 GB needed at FP16 + 8,192 ctx — overshoots effective VRAM by 49%. Drop quant or move to a larger build.
PaliGemma 2 10B	10B	BF16	26 GB	8,192	No measured row yet	~26.0 GB needed at BF16 + 8,192 ctx — overshoots effective VRAM by 86%. Drop quant or move to a larger build.
Mellum2 12B-A2.5B	12B	Q4_K_M	14.5 GB	8,192	No measured row yet	~14.5 GB needed at Q4_K_M + 8,192 ctx — overshoots effective VRAM by 3%. Drop quant or move to a larger build.
Baichuan 4 13B	13B	Q4_K_M	14.7 GB	8,192	No measured row yet	~14.7 GB needed at Q4_K_M + 8,192 ctx — overshoots effective VRAM by 5%. Drop quant or move to a larger build.
GLM-4V 9B	14B	Q4_K_M	15.9 GB	8,192	No measured row yet	~15.9 GB needed at Q4_K_M + 8,192 ctx — overshoots effective VRAM by 13%. Drop quant or move to a larger build.
Phi-4 14B	14B	Q4_K_M	15.8 GB	8,192	No measured row yet	~15.8 GB needed at Q4_K_M + 8,192 ctx — overshoots effective VRAM by 13%. Drop quant or move to a larger build.
Qwen 2.5 14B Instruct	14B	Q4_K_M	16.4 GB	8,192	No measured row yet	~16.4 GB needed at Q4_K_M + 8,192 ctx — overshoots effective VRAM by 17%. Drop quant or move to a larger build.
Qwen 2.5 Coder 14B Instruct	14B	Q4_K_M	15.8 GB	8,192	No measured row yet	~15.8 GB needed at Q4_K_M + 8,192 ctx — overshoots effective VRAM by 13%. Drop quant or move to a larger build.
DeepSeek R1 Distill Qwen 14B	14B	Q4_K_M	15.8 GB	8,192	No measured row yet	~15.8 GB needed at Q4_K_M + 8,192 ctx — overshoots effective VRAM by 13%. Drop quant or move to a larger build.
Phi-4 Multimodal	14B	Q4_K_M	16.5 GB	8,192	No measured row yet	~16.5 GB needed at Q4_K_M + 8,192 ctx — overshoots effective VRAM by 18%. Drop quant or move to a larger build.
Ministral 3 14B	14B	Q4_K_M	16.6 GB	8,192	No measured row yet	~16.6 GB needed at Q4_K_M + 8,192 ctx — overshoots effective VRAM by 18%. Drop quant or move to a larger build.
StarCoder 2 15B	15B	Q4_K_M	17 GB	8,192	No measured row yet	~17.0 GB needed at Q4_K_M + 8,192 ctx — overshoots effective VRAM by 21%. Drop quant or move to a larger build.
DeepSeek V2 Lite Chat	16B	Q4_K_M	16.9 GB	8,192	No measured row yet	~16.9 GB needed at Q4_K_M + 8,192 ctx — overshoots effective VRAM by 21%. Drop quant or move to a larger build.
DeepSeek V3 Lite (16B MoE)	16B	Q4_K_M	18 GB	8,192	No measured row yet	~18.0 GB needed at Q4_K_M + 8,192 ctx — overshoots effective VRAM by 28%. Drop quant or move to a larger build.
DeepSeek Coder V2 Lite (16B)	16B	Q4_K_M	18 GB	8,192	No measured row yet	~18.0 GB needed at Q4_K_M + 8,192 ctx — overshoots effective VRAM by 28%. Drop quant or move to a larger build.

Multi-GPU buying guide →

NVLink vs PCIe, tensor- vs pipeline-parallel, mixed-card honesty.

Hardware combinations →

Curated multi-GPU / cluster setups with effective-VRAM math.

Setup path-finder →

OS + runtime install commands for your stack.

Compatibility matrix →

Runtime × OS × hardware support truth table.

Shopping a full build instead of a single card?

If you're sizing a fresh AI build (not just a card to drop into an existing system), the build-budget walkthroughs cover the whole BOM honestly: AI PC build under $1,000 or AI PC build under $2,000 cover the realistic 2026 budget tiers.

Vertical-fit shopping? AI PC for students covers the budget + portability tradeoffs; AI PC for developers covers the coding workflow specifics; AI PC for small business covers the document-RAG / always-on machine.

Form-factor first? See best laptop for local AI, best Mac for local AI, best mini PC for local AI, or best used GPU for local AI.