Production reference / Forlex
Specialized models.
A working legal product.
Forlex is our first vertical. It brings legal research, document work and source checks into a product used by lawyers. Its LIVIA models run on Grid.
Reported adoption / August 2026
240,000
Monthly active users reported for Forlex in August 2026.
An adoption reference. Workflow quality and acceptance are evaluated separately for each engagement.
Published infrastructure case / Forlex
The migration in numbers.
Forlex’s published case study describes its move to Amazon EC2 P6-B200 and related AWS and NVIDIA tools. AWS reports these changes in the Forlex infrastructure workload.
Read the AWS case study- Lower operating costs
- ~40%
- Lower time to first token
- ~67%
- More tokens processed per minute
- 170%
Reported for this Forlex migration. The published case does not state a measurement period or sample size. Define the baseline and acceptance criteria for your own workload before making a production decision.
Prepare your evaluation planForlex-specific controls
Review at the points that matter.
These layers describe the Forlex product. Other enterprise workflows need their own scoped controls and acceptance tests.
01 / Input review
OpenDetector
Screens inputs for prompt injection before they reach the model.
02 / Output review
OpenJuris
Checks generated outputs for unsupported citations and hallucinations.
03 / Document faithfulness
OpenContracts
Verifies generated text against the source document.
Define the proof for your workflow.
Agree the baseline, evaluation set, quality threshold, human review and acceptance criterion with your team. Then assess response time and total cost per accepted result.
Prepare your evaluation plan