🧪 Lab
Model Benchmark Registry
Head-to-head model benchmarks for task delegation. Every entry compares two LLMs on a real task type — so we know which model to route to which job. Data: DeepSeek V4 Flash, Groq Llama 3.3 70B, Cerebras gpt-oss-120b, Mistral Large.
| # | Task Type | Model A | Model B | Winner Model | Metrics | Why Won | Date |
|---|
📝 Note: Entries are added after each benchmark run. Metrics compare speed, quality, and output completeness. Green badge = winning model for that task type.