Research · Open
Open models and tools.
What we release so others can check our work and build on it.
border.flowx.ai
Visit border LLM input & output checks with signed evidence
Open detectors on Hugging Face: prompt injection, groundedness, moderation, toxicity, bias, PII and more.
Methodology
How we build & evaluate models
Human-reviewed labels, deterministic models in immutable versions, and a frozen test set every version must clear.
Read the methodology →Open modelsOpen-weight models
Small Apache-2.0 models FlowX trains and releases on Hugging Face, with each model card's benchmarks.
See the models →Planned
GAUGE harness
The harness behind GAUGE, with synthetic example cases, once the benchmark has launched.
After launch