MiaAI LAB Contact

Local AI,
tested in the open.

I run open-weight and frontier models side by side, same prompt, one shot each, and publish every raw output. See what they build, and where they break.

Local AI Models
Building with AIPushing Local AI ForwardOpen-Weight ModelsNVIDIA DGX SparksPremium Local AI RecipesQuantization Deep DivesModel Comparisons & RecommendationsBenchmarksNo Cherry-PickingTips & TricksAgentic Workflows

New recipe · two DGX Sparks

Qwen3.8 Flash Next, now on two DGX Sparks.

Powered by TensorFold

A million tokens of context, images and video, and up to sixteen conversations at once, served from two Sparks on your desk with TensorFold's Zig engine. Run NVIDIA's NVFP4, or the faster INT4-AutoRound at 89.3 tok/s. Free, with every patch and every number published.

One request
67.4tok/s
Sixteen at once
390.1tok/s combined
Context
1Mtokens a request
Prefill
2,575tok/s, 32k prompt

Found the needle in a 998,986-token prompt.

See the full recipe Get it on GitHub
See all recipes
TensorFold · Qwen3.8-Flash-Next · 2× DGX Spark
67.4 tok/s Sample text, played at the measured speed

New research · In validation

Engine-Aware Quantization, built for the engine that runs it.

Lower error
41%on what the model writes
Lower error
54%on agent decisions
Speed
≈about the same
Discover EAQ
Mia, line-art portrait
Mia@MiaAI_lab
github.com/MiaAI-Lab

About

Hi, I'm Mia! I'm publicly developing the frontier of local and secure intelligence.

I'm a developer who got hooked on running AI locally. Most days I'm deploying and optimizing local AI models to run the BEST way possible for NVIDIA DGX Sparks and other devices I have in my lab.

From GLM, DeepSeek, Qwen, and others, I deploy them all, and make them work at their best state possible on the available hardware.

I'm also comparing and benchmarking all closed frontier AI models, including Claude, GPT, Grok and others, to see how they compare to the best local has to offer.

Models offered

The frontier of local intelligence: free, open-source recipes for NVIDIA DGX Sparks and RTX GPUs.

Each recipe runs on my own Sparks, and the benchmarks, patches and launch scripts are all published on GitHub. Clone one and serve a frontier-class model on your own hardware.

GitHub stars
6,968
Forks
1,055
Contributors
156

Across 28 recipes and sparkDash

Mia’s Favorite

Needs 2 Sparks

Vision

GLM 5.3 Flash

TensorFold Zig · EXL3 4-bit · DFlash2

47.8 tok/s

Decode speed, one stream

Peak
87.3 tok/s ×4 streams
Prefill
1,801 tok/s 32k prompt
Context
1M · a ~2.1M shared pool

0 stars0 forks1 contributor

Last updated

See the full recipe

Mia’s Favorite

Needs 4 Sparks

Vision

DeepSeek V4.1 Flash 1017

TensorFold Zig · FP8 + MXFP4 · DSpark

65.8 tok/s

Decode speed, one stream

Peak
165.3 tok/s ×8 streams
Prefill
4,127.2 tok/s 32k prompt
Context
1M · up to 16 at once

59 stars19 forks11 contributors

Last updated

See the full recipe

Needs 1 Spark

Vision

Qwen3.8 Flash Next

vLLM · NVFP4

48.7 tok/s

Decode speed, one stream

Peak
163 tok/s ×8 streams
Prefill
1,769 tok/s 32k prompt
Context
262k · 512k with YaRN

627 stars89 forks8 contributors

Last updated

Get the free recipe
See all 28 models

Plus 25 more recipes for DGX Sparks and NVIDIA GPUs, and sparkDash, the dashboard I benchmark with.

Install with an AI agent

Install this model

Paste this into an AI coding agent that can run commands on the machine that will serve the model (the head node, for a Spark cluster), such as Claude Code or Codex. It reads the recipe, adapts it to your setup, starts the server and checks that it works.


        

          
          
        

      

    
  








  

    

      

        
New in the Lab · Projects

        
sparkDash 2.70, every Spark in one window.

        
My open-source dashboard for NVIDIA DGX Sparks and NVIDIA GPU machines, and the benchmark behind every speed on my recipe pages. New in 2.70: alert notifications to ntfy, Discord or Slack, 30 days of metrics history, a cost compare page, and NVIDIA Flare support.

        
    
          
  • Live GPU, unified memory, storage, network and LLM metrics for one Spark or a whole fleet
    • 
          
  • Decode, Prefill, Quality and the new Tool Eval Bench, with every run kept to compare
    • 
          
  • A new shell: sidebar, command palette, four themes, and start or stop your models from it
    • 
        

        

          See sparkDash 
          GitHub 
        

      

      
    

  







  

    

      
      

        
New in the Lab · Grok Bots

        
Grok bots that work while you build.

        
Small Grok bots I build for my own lab and publish for free. Each one does a single job in the background, so you can stay on the work that matters.

        Meet the Grok Bots 
      

    

  







  

    

      
        
      
      

        
The book

        
From Zero to Seen

        
Get monetized on X

        
How I grew my X account from 200 to over 41,000 followers in about four months, the honest way, and how to get monetized on X.

        

          Get the book 
          About the book 
        

      

    

  







  

    

      

        
Support & contact

        
How to support the frontier of local intelligence.

      

      
Everything here is independent. Donations pay for GPU time, API credits and the coffee behind late-night runs.

    


    

      
      

        
Pushing Local AI Forward

        

        
          
          Become a sponsor of Mia's AI Lab on GitHubMonthly or one-time, through GitHub Sponsors
          
        
        
          
          Support Mia on Ko‑fiA quick tip toward the lab, through Ko‑fi
          
        
        

        
or donate with crypto

        
Donate with crypto through NOWPayments. Pick an amount and a coin, and you'll finish the payment on their secure checkout.

        

          

            Amount (USD)
            

              
              
              
              
              
            

          

          

            
            
          

          

          
          

          
Minimum $10 · network fees paid by sender

        

      


      
      

        
Contact

        
A question, a model you want tested, a collab idea? I read every message.

        

          

            

            

          

          

            
            
          

          

          
          

          

            
            

          

        

        
or find me on

        

          @MiaAI_lab
          GitHub
          Hugging Face