Send one prompt to many AI models side by side and compare the answers
The AI Lab runs one topic past several AI models at once so you can compare their answers before you commit to a provider. Choose the number of participants, name each one, assign a model, and add optional instructions. Available models include Claude Fable 5.1, GPT-6 Astra, Grok 4.6, Gemini 3.1, Mistral and DeepSeek V4.
Set the turns per participant, enter the topic or prompt, and start. Because the session is turn-based, each model sees and responds to the others, which shows how they handle disagreement, follow-up and correction, not just a first answer. You can also load a saved agent as a participant with its full prompt included.
Use it for model selection, prompt testing across providers, and evaluating tone and reasoning for a specific task. Stop at any point, save the session for reference, or clear the log and run a new topic. A model that errors or times out is excluded so the rest can continue.
What you can do
- Any number of participants, each with a name, a model and optional instructions
- Models including Claude Fable 5.1, GPT-6 Astra, Grok 4.6, Gemini 3.1, Mistral and DeepSeek V4
- Turn-based conversation so you see how each model responds to the others
- Load a saved companion or agent as a participant with its prompt included
- Set turns per participant, stop at any time, and save the session
How it works
- Choose how many AIs take part and give each one a name.
- Assign a model to each participant and add optional instructions or load a companion.
- Set turns per participant, enter the topic, and start the conversation.
- Read the exchange as it unfolds, then save the session or clear and run again.
Frequently asked questions
Which models can I compare?
The picker includes 15+ models across providers, among them Claude Fable 5.1, GPT-6 Astra, Grok 4.6, Gemini 3.1, Mistral and DeepSeek V4. Each participant in a session can run a different model.
Do the models see each other's answers?
Yes. The session is turn-based. Each participant sees the conversation so far and responds to it, so you can compare not only first answers but how each model handles disagreement and follow-up.
Can I use my own companion or agent?
Yes. Load a saved companion or agent as a participant. Its source prompt is included on every turn, which raises the token cost, and you can add short conversation-specific instructions on top.
What happens if a model fails?
A participant that errors or does not respond within the timeout loses its turn and is excluded, with a note in the log. The conversation continues as long as at least two working models remain.
Does it cost tokens?
Yes. Each turn is billed at the normal rate for the model that produces it, and including an imported companion's full prompt on every turn adds to the cost of that participant.