---
title: README
emoji: 🔁
colorFrom: blue
colorTo: indigo
sdk: static
pinned: false
---

# EnvLoop Research

EnvLoop builds environments and evaluation for agents. EnvLoop Research publishes the models, benchmarks and code
behind that work, with pre-registered evaluation and the data needed to check the numbers.
Homepage: [www.envloop.ai](https://www.envloop.ai) · Contact: [research@envloop.ai](mailto:research@envloop.ai)

## Releases

**Pev** — a calibrated fast-decision model for personal agents with long-term memory, and the benchmark it was
evaluated on.

- Paper (PDF): [pev.pdf](https://github.com/EnvLoop/Pev/blob/main/docs/tech-report/pev.pdf)
- Model: [EnvLoop/Pev-27B-LoRA](https://huggingface.co/EnvLoop/Pev-27B-LoRA) (gated, non-commercial research)
- Benchmark: [EnvLoop/Pev-Bench](https://huggingface.co/datasets/EnvLoop/Pev-Bench) (gated, non-commercial research)
- Leaderboard: [EnvLoop/Pev-Leaderboard](https://huggingface.co/spaces/EnvLoop/Pev-Leaderboard)
- Code and pre-registration: [github.com/EnvLoop/Pev](https://github.com/EnvLoop/Pev)