Elorian Glimpse: visual-reasoning model with a 'visual scratch space'
Elorian's first public model re-examines images and video mid-reasoning (zoom, crop, extract frames, run code) instead of tokenizing once, beating GPT-6 Astra on chart reasoning and video captioning per its own evals.
Why it matters
A few-hundred-GPU startup claims frontier-beating visual reasoning at $0.60/$3 per 1M tokens (roughly 94% below GPT-6 Astra) by making the model keep looking at the source material during inference.
What you could build with it
A robotics startup could pipe camera feeds through Glimpse's API to interpret cluttered scenes and read scientific charts mid-reasoning, getting a planner that costs cents per episode instead of dollars on frontier models.
Does it hold up?
Too early to judge — research preview with live API and waitlisted chat; only vendor-published benchmarks and pricing tables exist so far, no independent evals.
Built with Elorian Glimpse
- Elorian launches Glimpse, a visual reasoning model that re-examines imagesarticle · Independent analysis of the launch: pricing math vs Astra/Opus/Sol, the 'Humanity's Sixth Sense' benchmark context, and why a 94% discount may not beat developer switching costs.
- Elorian prices Glimpse 94% below GPT-6 Astraarticle · Pricing analysis with Elorian's Table 1 ($0.60/M input vs $10 Astra) and an episode-cost comparison on a LIBERO robot-planning task.
Learn more
First spotted on article: source.
More AI models
OpenAI publishes 722 AI-generated math manuscripts from an unreleased modelOpenAI released 722 mathematical manuscripts in 372 result families to GitHub under Apache-2.0, produced by an…models · JEV 0.74StepFun Step-5-Preview — open BF16 weights releasedStepFun published the full BF16 weights for its flagship Step-5-Preview agentic model (600B-parameter sparse MoE, 27B…models · JEV 0.73ElevenLabs Eleven v4 + v4 Turbo: new emotive TTS architecture with a 100ms real-time variantElevenLabs released Eleven v4 and Eleven v4 Turbo on September 28: a new text-to-speech architecture with inline…models · JEV 0.72Mistral Large 4 (Le Chonk): 1.05T-param open-weight MoEMistral's new flagship: a 1.05T-parameter mixture-of-experts multimodal model with 49B active params and 1M-token…models · JEV 0.72VoiceStudio: open-source, fully-local ElevenLabs alternative with voice cloning, dubbing and dictation in 646 languagesVoiceStudio is an open-source, fully-local voice AI studio: voice cloning from a 3-second sample, voice design, video…models · JEV 0.71NVIDIA Nemotron-pretraining-Indic-v1: 2.8T-token open Indic pretraining corpusNVIDIA with IIT Madras's AI4Bharat released a ~2.8 trillion-token pretraining corpus covering 22 Indian languages plus…models · JEV 0.69
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.