byokit.

REAL INFERENCE / YOUR CURRENT SETTINGS

Does this model
work for you?

Already verified? Test it again. More contributors help us understand speed and variability. Use your own key, then share your measurement on GitHub to build the community evidence.

We never save your API key. It stays in this page’s memory and is sent directly to the selected provider. It is not saved in browser storage, included in records, or sent to byokit. Refreshing or leaving the page clears it.
Local model

Enter your connection settings. Saving sends no requests.
Model verification

Test with your key. Share what you learn.

SELECTED MODEL

Loading catalog…

Provider docs ↗

Connection

Calculating estimated cost…

Use a key for the selected provider. A small inference call may use your credits.

Request goes directly to

Speed benchmark

Prompt and output limit

The standard benchmark uses this fixed counting prompt and 128 output tokens. Custom settings still verify availability but are excluded from the standard speed median. Speed is estimated from two non-streaming calls; actual time to first token is not measured.

Copy this API call

This matches the selected endpoint, model, prompt, and output limit. Examples run in Node.js, Python, or your terminal and use the API_KEY environment variable. Your entered key is never included. 


        

          SDK calls return an OpenAI-compatible completion; cURL returns the provider's native
          response. Making this request may use credits; displaying or copying it makes no request.
        

      

      

        
      

      

        Earn a Model Verifier badge. A successful hosted-model call plus a
        published GitHub report earns your badge.
      

    

    

      No call has been made.
    

    
    

    

      
    

    
  

  
  

  
  

  
  
  

    How sharing works
    

      After you submit your GitHub issue, we automatically check whether its reports are live on
      this site. “Refresh report status” checks again; it never calls models or spends API credits.
    

  

  

  


      

        

          

            
PLAYGROUND / TEST ANY API QUERY

            
Query and test models across APIs.

          

          

            Filter models across protocols or providers with JavaScript.
Select models to
            verify in the shared drawer.
          

        

        

          

            JavaScriptbyokit is already imported
          

          

            

              
            

            

              CONSOLE OUTPUT
              
Your console output will appear here.

            

          

          

            ⌘ / Ctrl + Enter
          

          
Ready.

          
        

        

          Queries run locally in an isolated sandbox. Select up to 10 models from the results, then
          verify them in the drawer. Selection alone sends no provider requests.
        

      

      

        

          

            
MODEL LIST / CALLABLE BY DEFAULT

            
Every speed measurement helps.

          

          
        

        

          Models supported by BYOKit’s text-generation adapters appear by default. Uncheck “Callable
          only” to browse all catalog models. Call support does not guarantee access for your
          account. Availability and speed are separate. Standard speed summaries use the median and
          middle 50% of published comparable measurements from the last 30 days, grouped by provider
          offering and benchmark version. Browser location, network, provider load, and output
          length still affect results. Batch checks verify availability; use Test speed for the
          two-call speed benchmark.
        

        
        

          What can I filter?
          

        

        

          

            

            0 models selected
            

              Select up to 10 models across providers. Enter each provider’s key in the verification
              drawer. Requests run one at a time without retries.
            

          

          

            
          

        

        

          

            
              Model list, metadata tags, and verification actions
            
            
              

                
                  Select
                
                
                  Provider / model
                  

                    
                    
                  

                		
                
                  Context
                  

                    
                  

                		
                
                  Price / 1M tokens
                  

                    
                  

                		
                
                  Speed / release
                  

                    
                  

                		
                
                  Input
                  

                    
                  

                		
                
                  Tags & capabilities
                  

                    
                  

                		
                
                  Status / verification
                  

                    
                    
                    
                  

                		
              

            
            
          

        

        

          
        

      

      

        

          

            
COMMUNITY / MODEL VERIFIERS

            
A badge for your contribution.

          

        

        

          Publish your first successful hosted-model report to earn the Model Verifier badge. Your
          badge stays as recognition of your contribution; model checks have their own freshness
          dates. Share it in your GitHub profile README or project.
        

        

          
Loading published contributors…

        

      

      

        

          

            
SUCCESSFUL CALLS

            
Recorded in this browser.

          

          

            
          

        

        

          Attempts are saved here with their outcome, model/API settings, date, latency, and
          available token usage. Keys, prompts, and generated text are never saved. Records describe
          a past call; another key or changed settings need a new check. After a catalog-hosted
          model succeeds, Share verification opens a prefilled GitHub issue. Submit it there; GitHub
          Actions commits your attributed community report and updates the website. Opening the form
          alone does not publish anything. Local-server checks stay in this browser.
        

        

        

        

          Contribute a verification record ↗