Browser Nav Challenge
RL environment for post-training small models on Brett Adcock's 30-step browser navigation challenge using DOM accessibility trees and Playwright
- Domain
- rl-env
- License
- unknown
- Published
- Feb 2026
Cite
Notes
Only stored in your browser.
Top score 16.7% by Claude Opus 4.5 - 4 models reporting (2 frontier)
Score history
4Top models
4
Related tools
1Implementations, trainers, datasets and scaffolds linked to this eval.
FAQ
- What is Browser Nav Challenge?
- RL environment for post-training small models on Brett Adcock's 30-step browser navigation challenge using DOM accessibility trees and Playwright
- What is the current top score on Browser Nav Challenge?
- The top reported score is 16.7% by Claude Opus 4.5, across 4 models reporting (2 from frontier labs).
- How can a model improve its Browser Nav Challenge score?
- Tools linked to Browser Nav Challenge on Sophon include NAV Challenge RL Env (Chakra Labs) - RL environments, datasets, and scaffolds that target this eval.
- What license is Browser Nav Challenge under?
- Browser Nav Challenge is available under unknown.