Independent SpatialBench fully saturated by Astra, author declares LLM vision solved
pbaylies · x · 2026-09-06
- spiceylemonade maintains SpatialBench, a spatial reasoning benchmark testing VLM tracing, 3D visualization, and cross-view reasoning.
- Previously every new model he sampled failed his test questions, but Astra kept getting them correct.
- After a full evaluation, Astra completely saturates the benchmark, and the author declares LLM vision effectively solved.
More from Models
- Hands-on GPT 6 Astra review: real work, no game demos — Rasmic · 2026-09-06
- Astra day-two impressions: chattier, strong spatial sense, less jargon than Fable 5 — bindureddy · 2026-09-06
- 200 tok/s on 8GB VRAM: dev benchmarks 6 small models for local AI — TheMoonMidas · 2026-09-06
- Humanize + GPT-5.5 solves 670/672 Lean-verified proofs, tops PutnamBench at 99.7% — songhan_mit · 2026-09-06
- "Claude Solved Navier–Stokes" Is Just a Rumor: No Paper, No CMI Submission, Say Fact-Checkers — johnseach · 2026-09-06
- Frontier AI models begin crossing the human baseline on SimpleBench — Bojackin_Around · 2026-09-06