Text, documents and agents
Versions, access and task-specific criteria.
Choose a model for a specific task
Start with the task, acceptable errors and available checks. These guides organise versions and scores documented on 14 October 2026; they are not a LearnIA testing campaign.
Read benchmarks
Published scores, metrics, margins and protocol.
Video without filming
Consistency, movement and cost per accepted shot.
Voice and transcription
Pronunciation, errors, latency and use cases.
Decisions with Jev
Typed decisions, calibration and ambiguous cases.
Prepare your own comparison
- Define a task and expected answer before testing.
- Keep identical inputs and record versions and settings.
- Check critical errors, rework and cost per usable result.