Astra follow-up: ~120 benchmark outputs published to back SOTA claim
legit_api · x · 2026-09-05
Following the claim that Astra topped VoxelBench, @legitapi published around 120 model outputs for public inspection to back the SOTA claim. The model's identity remains undisclosed.
Related event: Mystery Model Astra Tops VoxelBench, Publishes 120 Sample Outputs(3 posts)→
More from Models
- Astra vs Fable: one follows your typo literally, the other quietly fixes it — bclavie · 2026-09-05
- GPT-6 Astra builds working Minecraft screen animating OpenAI logo into '6 Astra' — Angaisb_ · 2026-09-05
- Polymarket opens book on whether OpenAI ships GPT Astra 6.1+ this year — Polymarket · 2026-09-05
- AA Small Model Leaderboard Update: Ling 3.0 Tiny Leads With Just 1.3B Active Params — Tall_Abrocoma_3533 · 2026-09-05
- Codex users report quotas draining much faster than usual — xiaohu · 2026-09-05
- AA Frontier Leaderboard Update: Qwen3.8-27B Among the Standouts — Tall_Abrocoma_3533 · 2026-09-05