Nex-N2.5 Pro plays Pokémon across hundreds of steps, testing real computer-use endurance
alifcoder · x · 2026-09-13
Most computer-use demos stop after one clean action, but Nex-N2.5 Pro reportedly kept playing Pokémon across hundreds of steps—reading the screen, navigating menus, battling, and recovering from mistakes. According to OpenRouter, the Nex-N2.5 family is an agentic model built around agentic coding in a visual feedback loop: exploring codebases, making multi-file changes, running commands, driving browser/desktop interfaces, and testing software from the user's perspective, diagnosing and revising when behavior diverges from intent. Both Nex-N2.5-Mini and Nex-N2.5-Pro are free on OpenRouter with a 262K context window.
Related event: Nex-N2.5 Pro Plays Pokémon for Hundreds of Consecutive Steps(2 posts)→
More from Models
- OpenAI claims its model found a Navier-Stokes blowup counterexample, cracking a Millennium Problem — victor_explore · 2026-09-13
- Reddit user calls out OpenAI models for botching basic percentage math — nutinahut · 2026-09-13
- "Open-source AI must win": OpenMed manifesto argues powerful intelligence can't stay in few hands — MaziyarPanahi · 2026-09-13
- Mathematician cracks Navier–Stokes with Codex, then OpenAI agents did it in 88 hours — rubenhassid · 2026-09-13
- Grok Bots coming to XChat: leak suggests bot DMs on X — nima_owji · 2026-09-13
- Third-party 10-dim eval shows DeepSeek V4.1 big reliability gains over V4-Pro-0813 — teortaxesTex · 2026-09-13