Elorian's Glimpse matches Argon on labels but costs 25x more than Gemma 4 31B
arnabdotorg · x · 2026-10-18
relanticai benchmarked Elorian's new Glimpse model on RealityBench physical-work tasks: it reads labels as well as Gemini 4 Argon, but shows gaps in fine-grained visual reasoning like tape measures and lag bolts. Their run cost $18 for a score Gemma 4 31B matches at $0.70. The scatter chart shows frontier models near human-rater quality (GPT-6 Astra, Claude Fable 5.1) clustering above $10, while cheap models like GLM-5.3 Flash and Gemma 4 31B stay under $1 with decent scores.
More from Models
- Ex-Microsoft exec Parakhin: after 2 days with AI advisors, I still add value — MParakhin · 2026-10-18
- $20 plan showdown: user finds Codex far less token-efficient than Claude — cneuralnetwork · 2026-10-18
- Dave Shapiro ditches ChatGPT's Dot agent over silent failures and opaque model routing — _AustinCalvert_ · 2026-10-18
- OpenAI releases 700+ AI-generated math papers, with claimed progress on Riemann hypothesis — PolarBearby · 2026-10-18
- Steering a single neuron only works for concepts with good pretraining representation — Sauers_ · 2026-10-18
- Every Grok Bot task ends with writing a prompt for local ChatGPT instead — lucasmeijer · 2026-10-18