Phi-3-mini on raw WebGPU
Runs entirely in your browser on 10 hand-written WGSL kernels. No server, no API calls, no data leaves this tab.
~2 GB
weights
~40 t/s
M2 Pro
f16
WebGPU
Not now
Preparing Zero-TVM
Starting…
0%
Details 

      

      

    

  


  
  

    

      
        
      
      

        
Phi-3-mini

        
10 WGSL kernels · paged KV · WebGPU

      

      

      
      
      Waiting
    


    

      

        

          
          
Chat with Phi-3-mini

          

            q4f16_1 · 3.8 B
            4 K context
            ~40 tok/s · M2 Pro
            On-device only
          

          
Running locally on your GPU through 10 hand-written WGSL kernels. Nothing leaves this tab — prompts, tokens and KV cache all stay in your browser.

          

            
            
            
            
          

        

        

      

      
    


    

      

        
        
        
      

      

        Enter to send · Shift+Enter for new line ·
        Zero TVM · 10 WGSL kernels · 228 dispatches/token