The Model Playground allows engineers to compare foundational models side-by-side, inspect streaming latency (TTFT), and evaluate cost efficiency.

Portal Page: /projects/[projectId]/model-playground

Features #

Side-by-Side Comparison: Run the same prompt simultaneously against GPT-4o, Claude 3.5 Sonnet, Gemini 2.5 Flash, and DeepSeek R1.
Latency & TTFT Metrics: Real-time adapter duration and time-to-first-token benchmarking.
Token Breakdown & Cost Estimation: Exact prompt tokens, completion tokens, and calculated cost in USD.
Code Export: Export your tested prompt and parameter settings directly to cURL, TypeScript, or Python code.

Next Steps #

Test autonomous agents: Agent Playground
Configure provider keys: BYOK & Credential Pools

Model Playground & Prompt Engineering

Features #

Next Steps #