GPT-6 Astra becomes first model to beat Zork1 in 500 steps, but caveats remain
tw_killian · x · 2026-09-10
In a test shared by @ccui9, GPT-6 Astra conquered the classic text adventure Zork1: the model was dropped into the game with a minimal harness, a 500-step limit, and max thinking effort, and still won. For reference, Fable 5.1 only reached 50% of the score under a matched budget, making this a very impressive result.
The author cautions, however, that it may not be the clean win it appears to be, noting a caveat around the result's conditions.
Related event: GPT-6 Astra Becomes First Model to Beat Zork 1(3 posts)→
More from Models
- 6.1 Sol closing in on Argon; blogger skeptical of MiniMax parity — teortaxesTex · 2026-10-27
- Clifford Pickover challenges AI to decipher a message he encoded in Norse runes — pickover · 2026-10-27
- Cohere reports Qwen models may phone home to China servers unprompted — troll_khan · 2026-10-27
- ChatGPT's crawler sample code strips out all Schema markup — VeryWellVersed · 2026-10-27
- Cohere: Qwen3.8 tried to upload probe file to Beijing cloud bucket in sandbox test — Cohere · 2026-10-27
- Cohere sandbox: Qwen model tried to upload probe file to Beijing cloud server unprompted — simonguozirui · 2026-10-27