Lane A
ModelRun the prompt to populate this lane.
Latency: -- Tokens: -- Model: --
First experimental surface
Model Lab
Run the same task across multiple models, compare what changes, and inspect clarity, reasoning, tone, latency, and review posture in one calm interface.
Prototype status: live streaming comparison is now available through the OpenRouter lane runner. Add one prompt. Choose three models. Read and compare outputs side by side.
Composer
Set the task once, route it to three model lanes.
Results
Compare outcomes side by side without losing the review context.
Lane B
ModelRun the prompt to populate this lane.
Latency: -- Tokens: -- Model: --
Lane C
ModelRun the prompt to populate this lane.
Latency: -- Tokens: -- Model: --
How to read the lab
What matters is not just the answer, but how the answer behaves.
Compare the same task
Keep the prompt stable so the model behavior is the thing that changes.
Read for reviewability
Ask whether the answer makes it easy to verify, edit, and safely reuse.
Watch the tradeoffs
Some models are faster, some clearer, some deeper. The best one depends on the job.
Keep it draft-first
Outputs should move toward trusted capability, not instant overconfidence.