Run Local AI - free tool from

Does it run locally?

From a phone, a Raspberry Pi or a Jetson edge AI board to a gaming PC, a Mac or a data-center GPU, see which AI models your hardware can run, how well and how fast. Then pick the app to run them in, or compare what to buy next.

Detect my hardware Find models for my hardware

Updated daily

New local AI models past 30 days

See all 4 new models this week →
  1. Nex N2.6 (321B-A18B) A+

    Nex-AGI · 321B (18.0B active) · Nov 10

    • Phone
    • Laptop
    • Gaming PC
  2. MiniMax H3 Image A+

    MiniMax · 33.1B · Nov 10

    • Phone
    • Laptop
    • Gaming PC
  3. Raon 1 nano D

    KRAFTON · 3.5B · Nov 10

    • Phone
    • Laptop
    • Gaming PC
  4. LongCat 2.5 (1.6T-A48B) A+

    Meituan · 1,780B (48.0B active) · Nov 9

    • Phone
    • Laptop
    • Gaming PC
  5. Kimi K3.1 Pro (2.8T) A+

    Moonshot AI · 2,780B (104.0B active) · Nov 2

    • Phone
    • Laptop
    • Gaming PC
  6. S5 nano D

    Maritaca AI · 4.7B · Nov 2

    • Phone
    • Laptop
    • Gaming PC
  7. NorthCUA Mini B

    beipo-ai · 35B (3.0B active) · Nov 1

    • Phone
    • Laptop
    • Gaming PC
  8. NorthCUA Nano C

    beipo-ai · 9.4B · Nov 1

    • Phone
    • Laptop
    • Gaming PC
  9. GLiDE 9B C

    Fastino · 7.9B · Nov 1

    • Phone
    • Laptop
    • Gaming PC

Explore

Find the right model, hardware and app for local AI

LLMs 622 language models Image models 61 diffusion models Video models 43 video models Decision models 70 System One models Hardware 753 chips and devices Apps 75 desktop and phone apps My hardware Detect it, get a full report Compare Up to 4 systems at once

Model Finder

Can your GPU or device run it?

Loading the model catalog…

How the estimate works

Three numbers decide it

01 · MemoryDoes it fit?Model weights at each quantization, plus the KV cache for your context length and runtime overhead, against the memory your GPU or device can actually give a model.
02 · BandwidthHow fast will it answer?Generating each token reads the weights once, so tokens per second follow memory bandwidth. Mixture-of-experts models only read their active experts.
03 · QualityIs it the best that fits?A curated quality tier for each model carries most of the score, so the top pick is the best model that fits, not the smallest one.

Example

Llama 3.3 70B on a 128 GB Mac

At Q6_K it needs about 69 GB. A Mac with 128 GB of unified memory lets a model use about 96 GB, so it fits, at roughly 6 tokens per second.

See the full breakdown

Popular checks

What people ask

62Good quality (B) · 62 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 71/100 (quality, fit, speed and memory use) sets the ranking.Llama 3.3 70B Instructon Apple M4 Max (128 GB) · Fits in memory72Great quality (A) · 72 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking.Qwen3 32Bon NVIDIA GeForce RTX 3090 · Tight fit70Great quality (A) · 70 of 100: how capable the model is at MXFP4, whatever it runs on. Match score 86/100 (quality, fit, speed and memory use) sets the ranking.gpt-oss 20Bon NVIDIA GeForce RTX 4060 Ti 16GB · Fits in memory78Great quality (A) · 78 of 100: how capable the model is at FP8, whatever it runs on. Match score 66/100 (quality, fit, speed and memory use) sets the ranking.FLUX.1 [dev]on NVIDIA GeForce RTX 4060 · Tight fit60Good quality (B) · 60 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 81/100 (quality, fit, speed and memory use) sets the ranking.Gemma 3 27B ITon NVIDIA GeForce RTX 4070 Ti SUPER · Tight fit77Great quality (A) · 77 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 86/100 (quality, fit, speed and memory use) sets the ranking.DeepSeek R1 (671B)on Apple M3 Ultra (512 GB) · Tight fit

From the team at EnableVue

Running AI on your product catalog?

EnableVue turns messy product data and thin web content into catalogs that rank in Google and get cited by AI answer engines. Expert-led services, AI-accelerated delivery.

Visit EnableVue

Was this page helpful?