Elorian's Glimpse claims frontier visual reasoning at 16x lower cost than Astra
thegautamkamath · x · 2026-10-17
Palo Alto startup Elorian launched Glimpse, its first "visual thinking" model for physical-world reasoning, targeting models' weakness on moving objects and changing viewpoints.
- Three core tasks: tracking identity through motion, reasoning from changed viewpoints, predicting physical consequences
- Reasons in visual space, revisiting internal visual representations while thinking; more grounded in objects and spatial relations
- Trained end-to-end as an agent with multimodal tools (code, image transforms, video) and can spawn coordinating subagents
- Claims frontier-level performance at or above GPT-6 Astra and Claude Opus 5.5 while being much smaller and 16x cheaper
- Open research preview via chat and API; aimed at robotics, science and creative work
More from Models
- KOL teases mystery model argon, says people are 'severely underestimating' it — iruletheworldmo · 2026-10-17
- Perplexity decider v1.2 lands #3 on Decision Index as longer training erodes backbone skills — antoine_chaffin · 2026-10-17
- Perplexity's 4B decider model distills from its 27B sibling using the same training recipe — antoine_chaffin · 2026-10-17
- Perplexity traces decider v1.2 regressions to longer training, mitigates via model souping — antoine_chaffin · 2026-10-17
- Perplexity details decider data hygiene: official splits only, deduped, 16.6% of one stage — antoine_chaffin · 2026-10-17
- Perplexity says rivals train on Decision Index benchmark sets, opts to follow for fair comparison — antoine_chaffin · 2026-10-17