Models Tasks Hardware Runtimes Benchmarks Deploy
100+ Verified Local Models

Run AI on Your Hardware

Compare models by task and quality, then generate the exact setup you need.

Popular:
Explore models
No data leaves your machine
12+model categories
8runtime profiles
1-clickdeploy recipes
LocalModels Match
Your hardware RTX 3090 · 24 GB VRAM
Excellent fit 94 fit score

ForgeCoder 22B

Code generation · agents · tool use

RecommendedQ4_K_M
Memory15.6 GB
Speed42–51 t/s
Context32K
02
Atlas Instruct 14BFaster general option
89
03
VisionLeaf 11BBest multimodal fit
86
Estimated speed 47 t/s
Memory headroom 7.1 GB
Language modelsVisionVoiceEmbeddingsImage generationAgentsRoboticsPrivate RAGLocal APIs Language modelsVisionVoiceEmbeddingsImage generationAgentsRoboticsPrivate RAGLocal APIs
Start with the work

What do you want AI to do?

Choose a workload and we’ll rank the strongest models that fit your machine, deployment, and licensing requirements.

Model explorer

Models matched to the way you work.

6 results
Hardware-aware by default

Know what will run before you download it.

Model size is only part of the equation. LocalModels.io accounts for quantization, runtime overhead, context memory, hardware acceleration, and practical performance.

  • 01 Select your actual machine
  • 02 Choose a task and target context
  • 03 Get the best artifact and runtime
Compatibility report MacBook Pro · 36 GB
Ready
91fit score
Memory fitExcellent
Expected speedGood
Context capacity32K
Setup complexityEasy
Recommended stack
Atlas 14BMLX · 4-bit · 32K context
Runtime directory

Choose how you want to run it.

Compare desktop simplicity, local APIs, production throughput, Apple optimization, multimodal support, and deployment complexity.

Fastest start

Ollama

Straightforward model installation and local APIs for applications, agents, and rapid prototyping.

macOSWindowsLinux
View runtime →
Best desktop UI

LM Studio

Discover, compare, chat, and serve local models through a polished desktop environment.

DesktopLocal APIGUI
View runtime →
Apple optimized

MLX

High-efficiency local inference workflows designed for Apple silicon and unified memory.

Apple siliconPythonCLI
View runtime →
Production serving

vLLM

High-throughput GPU serving for multi-user APIs, production workloads, and larger systems.

GPU serversAPIScale
View runtime →
Benchmarks with context

Measured on real hardware.

Compare quality, speed, memory, latency, and reliability using complete, reproducible configurations—not isolated numbers.

Explore benchmarks
Generation speed
ForgeCoder 22B
48
Atlas 14B
43
VisionLeaf 11B
37
ReasonCore 32B
24
0Tokens per second50
From discovery to deployment

Go from model page to working infrastructure.

Generate commands, runtime configurations, Docker deployments, private-server setups, and production-ready handoff packages.

Local installDockerPrivate serverMulti-GPUKubernetesEdge
localmodels deploy
# Generated for RTX 3090 · 24 GB
$ localmodels deploy atlas-14b \
  --runtime ollama \
  --quantization q5_k_m \
  --context 32768 \
  --bind localhost

✓ Hardware validated
✓ Artifact verified
✓ Estimated memory: 13.8 GB
Ready at http://localhost:11434
LocalModels Enterprise

Operate approved models on infrastructure you control.

Evaluate, approve, deploy, and track public and private models across teams, private networks, and air-gapped environments.

Private model registry Approval workflows License tracking Deployment inventory
Talk to us
Start with what you already own

Your next model may already fit on your hardware.

Tell LocalModels.io what you have and what you want AI to do.

Browse all models