Azure Anthropic0.0 s
versus / best vs best
Same question.
Every model.
You call the winner.
Pick 2 to 39 models from four clouds, ask once, and watch every answer stream side by side from the same context. Fastest, clearest, most accurate: you decide.
you
4 models · 1 requestwaiting
Azure Foundry0.0 s
Google Gemini0.0 s
Azure Foundry0.0 s
demo replay · in Versus you pick the models and the answers are live
Claude Opus 5.5
GPT 6.1 Sol![]()
any two of 39 · or up to all 39 at once · mix clouds freely
01 / one stream
One request.
Every answer at once.
Versus does not call the models one after another. It opens a single event stream and fans out to every model you picked in parallel, across Azure, Google and AWS.
- 01
init
The stream opens and names every model in the battle.
- 02
start · chunk
Each model streams on its own, every chunk tagged with its model id, so the page drops it into the right column.
- 03
complete · error
Each model ends on its own: a finished answer, or an error that stays in that lane.
- 04
done
The whole round is saved once, with every answer and its time, as one turn of your battle thread.
simplified event log · a failed model gets a generic message, never the provider's error text
02 / rules
Rules of the arena.
A comparison only means something if every model gets the same chance. These are enforced in the backend, not promised in copy.
- 01
Same context
Retrieval runs once per question, and every model reads the same five chunks. Nobody gets a better brief.
- 02
Same start
Every selected model is called at the same moment and streams into one response, tagged by model id.
- 03
Same clock
Each model races its own 120 s timeout. Slow is allowed; hanging forever is not.
- 04
No sore losers
A model that fails sends its own error event, with a generic message. The others keep streaming.
- 05
One count
A battle counts once against your 3 questions per 12 hours, whether it runs 2 models or 36.
- 06
Battles have threads
Follow-ups carry the earlier rounds, and every round is saved with its thread, so you can come back to it.
03 / roster
Pick your lineup.
39 models, four clouds.
Try it: tap models to build a battle. The same models, grouped the same way, are in the Versus picker.
Azure Foundry16 models
Azure Anthropic9 models
Google Gemini4 models
AWS Bedrock10 models
3 models in this battle · 1 question, 1 count against your limit
Start this battleHow Versus runs a model battle.
One question, every model you picked, all streaming at once. Click any step to see what runs there and where it lives in the code.
04 / your call
Settle the
argument.
Stop guessing which model is best for your question. Put 2 to 39 of them in the same arena and read the answers side by side.
Start a battle