Five-round vision benchmark: GPT-6 Astra vs Hy4 on detection, video and game tasks
IlovedamemesUwU · reddit · 2026-11-11
A blogger ran five same-brief challenges comparing Hy4's newly upgraded vision against GPT-6 Astra, with GPT-6.1 Sol authoring the prompts.
- Hidden keys: both found them in nearly identical time (2m15s vs 138.57s) — a tie.
- Lipstick trap test: Astra 6 misidentified a lookalike battery as lipstick; Hy4 re-checked the whole image and honestly reported no fourth lipstick existed.
- One-line prompt to 10s video: Astra 6 hit a tool-install hiccup, retried, finished in 15 min, staying closer to the brief but stiff; Hy4 succeeded first try in 22 min with its own interpretive layer.
- Poster to motion ad: both good; Hy4's light-based transitions gave stronger overall motion.
- Hand-drawn sketch to game: author's favorite round (post truncates here).
Overall Hy4 impressed on detail and refusing to force answers; Astra 6 led on speed and prompt fidelity.
More from Models
- Gemini 4 Carbon launches with $2/$10 intro pricing; Pydantic AI ships day-1 support — samuelcolvin · 2026-11-11
- Google open-sources Harness², an agent that rewrites its own setup per task — Arindam_1729 · 2026-11-11
- Dev dumps Codex desktop app over quality issues, falls back to codex-cli — tekbog · 2026-11-11
- Why do old models get dumber before launches? Lab party, says joking X post — glcst · 2026-11-11
- Grok Bot now detects hidden prompt injections, warns users and refuses to follow them — XFreeze · 2026-11-11
- Shanghai Institute Open-Sources Nex-N2.6, a 321B Multimodal Agentic Model — victormustar · 2026-11-11