CAAP models.x402.wtf Solana AI Model Kit

Chat and code benchmarks

Compare any OpenAI-compatible model, Anthropic, Gemini, or local mock runs.

Each run records model outputs, latency, token counts, code execution results, and a shareable X payload without storing provider keys.

Results Training builder
Solana AI Model Kit pipeline

Realtime results

Benchmarks

No active run Run a chat or code arena to populate the benchmark table.

model-arena-events

Waiting for a run...

Recent runs

Local session

x402 model launch surface

Train from local files, publish to Hugging Face, and hand the model to CAAP/1.0.

The kit turns PDFs, JSONL, CSV, notebooks, markdown, YAML, parquet, and image sidecars into a reviewed SFT dataset, optional LoRA adapter, and registry-ready payload.

Open register.x402.wtf Consolidation map $CLAWD coin x402.wtf
Solana AI Model Kit pipeline

clawd-model-kit


          


          

            

              

                
Artifact map

                
ai-training is the canonical model workspace.

                
The July 4 consolidation moved legacy root outputs, fresh NVIDIA CPT/SFT checkpoints, and release notes into one model-kit-visible lane.

                

                  Read article
                  View models
                

              

              Animated ai-training consolidation flow
            


            

              
Flow

              
One-shot path

              
    
                
  1. CollectDrop source files into `ai-training/data/incoming`.
    2. 
                
  2. BuildCreate SFT JSONL, splits, manifest, and dataset card.
    3. 
                
  3. TrainRun a local LoRA dry-run or launch a guarded HF Job.
    4. 
                
  4. RegisterSend CAAP/1.0 metadata with the Constitution hash commitment.
    5. 
              

            


            

              
Harness

              
Constitution gate

              
Model-kit commands verify the root Constitution and hash-attested on-chain laws before ingest, train, upload, or register workflows.

              

                three_laws fa1c36ab...31c2bc
              

              
Current three-laws hash: fa1c36ab...31c2bc

            


            

              
Deploy

              
Vercel plus Render

              

                cd ai-training/model-kit
                npm run build
                vercel deploy --prod
                render blueprint launch ai-training/model-kit/render.yaml
              

            

          


          

            

              
Published datasets

              
Training lanes

            

            

          


          

            

              
Models

              
Adapters and foundation lanes

            

            

          


          

            

              
Jobs

              
Training status