Elorian, the Palo Alto visual-AI startup whose seed round included Nvidia, according to TechCrunch, released its first model, Glimpse, on 16 October 2026 and priced it at $0.60 per million input tokens and $3.00 per million output tokens. Elorian says “both rates are 94% lower than GPT-6 Astra’s, 85% lower than Claude Opus 5.5’s, 70% lower than GPT-6.1 Sol’s, and 20% lower than Gemini 3.8 Flash’s introductory rates.”
On its own benchmark table, Elorian says “Glimpse exceeds Astra on four evaluations spanning scientific chart reasoning, open-vocabulary counting, short-form video captioning, and manipulation-event understanding, and matches it on embodied reasoning.” The same table shows Glimpse below GPT-6 Astra on six other evaluations.
Business Pill · COST PER TASK
A one-minute explainer of cost per task: paying for the job done rather than for each word. It teaches the general idea only and says nothing about any company in this story.
The key insight: As we read it, Elorian is competing on price per task as much as on scores: it lists Glimpse at $0.60 per million input tokens against $10.00 for GPT-6 Astra, says its rates are 94% lower, and in one robot-planning example puts Glimpse’s cost at $0.035 for the episode against $0.249 for Astra.
Glimpse is a research preview available through an API and a chat interface, Elorian says. Its post says “We developed Glimpse as a small team with only a few hundred GPUs, post-trained on an open weights base model.”
TechCrunch, which says it exclusively learned of the launch, says the chat interface is waitlist-only for now, while Elorian’s post says Glimpse “is available through API and our chat UI today.” The two accounts differ on chat access.
Elorian calls Glimpse “a new category of visual thinkingmodel: a system that reasons about the physical world from what it sees.” Its post says Glimpse can zoom into a region, watch a short interval at a denser frame rate, or track one feature across time, which it calls the “visual scratch space”.
Elorian’s Table 1 (16 October 2026) lists uncached input prices per million tokens of $0.60 for Glimpse against $10.00 for GPT-6 Astra, $4.00 for Claude Opus 5.5, $2.00 for GPT-6.1 Sol and $0.75 introductory for Gemini 3.8 Flash.
What Was Posted on X
Today we’re thrilled to announce the release of our first frontier visual reasoning model
Elorian’s own account announced Glimpse on 16 October 2026 as its first frontier visual reasoning model; the figures in this piece come from the launch post the announcement points to.
๐จHOLY MOLY. Nvidia-backed AI startup Elorian has released its first model, Glimpse, claiming it beats GPT-6 Astra on on four benchmarks covering robotic manipulation events, scientific charts, object counting and short-form video captioning.
Trained by a small team on a few hundred GPUs, Glimpse uses a ‘visual scratch space’ to repeatedly inspect images, zoom into details, extract video frames and execute code during reasoning.
The API costs $0.60 per million input tokens and $3 per million output tokens, about 17x cheaper than Astra’s standard rates.
Elorian previously raised $55 million at a $300 million valuation, with backing from Nvidia and Google Chief Scientist Jeff Dean. Glimpse is launching as a research preview, with API access live and chat access waitlisted.
An investor account summarised the launch: Glimpse ahead of GPT-6 Astra on four benchmarks, a few hundred GPUs, API pricing of $0.60 and $3 per million tokens, and the $55 million seed with Nvidia and Jeff Dean. Its ‘about 17x cheaper’ figure is the poster’s arithmetic on Astra’s standard rates.
The Price Against Frontier Models
Elorian’s Table 1 lists uncached input prices per million tokens of $0.60 for Glimpse, $10.00 for GPT-6 Astra, $4.00 for Claude Opus 5.5, $2.00 for GPT-6.1 Sol and $0.75 introductory for Gemini 3.8 Flash. Output prices are $3.00, $50.00, $20.00, $10.00 and $3.75 introductory respectively.
In one robotics example in the post, Glimpse and GPT-6 Astra each act as the planner for the same robot policy on the same LIBERO task. Elorian says both complete it: “Glimpse’s planning costs $0.035 for the episode; Astra’s costs $0.249.”
What the Benchmarks Show
On WGO Bench, for end-to-end manipulation events, Elorian lists Glimpse at 42.70 against 39.42 for GPT-6 Astra. On CharXiv-R, for scientific chart reasoning, it lists 92.90 against 92.10, and on UGC-VideoCap, for short-form video captioning, 82.66 against 80.32.
On FSC-147, open-vocabulary counting, the table lists Glimpse at 83.36 against 82.02, and on ERQA, embodied spatial reasoning, both at 83.75. GPT-6 Astra is ahead on HSS (53.6 against 46.36), Video-MME v2 (64.06 against 56.68), CountQA, DREAM-1K, TOMATO and BabyVision.
Elorian notes the conditions: “Unless otherwise noted, all model results use the highest thinking effort and native agent harness”. HSS is a benchmark Elorian built with Scale AI and announced on 8 October, according to its blog.
Who Is Behind Elorian
TechCrunch reports that Elorian raised a $55 million seed round at a $300 million valuation before launching a product, in a round co-led by Striker Venture Partners, Menlo Ventures and Altimeter Capital, with participation from Nvidia and Google chief scientist Jeff Dean.
Its chief executive is Andrew Dai, a former Google DeepMind researcher, according to TechCrunch. “Humans can look at Ikea diagrams, with no text, and assemble Ikea furniture, right? Why can’t AI models?” Dai told TechCrunch.
As we read it, the comparison is narrow by design: Elorian says Glimpse exceeds Astra on four evaluations and matches it on a fifth, while the same table shows Astra ahead on six others, including HSS and Video-MME v2.
As we read it, the build is the other half of the claim: Elorian says a small team developed Glimpse with only a few hundred GPUs by post-training an open weights base model.
As we read it, the target users are named: TechCrunch reports the initial targets are companies sitting on lots of video, including robotics firms, scientific labs, media and creative companies.
Elorian, launch post (16 October 2026)
“On multimodal benchmarks, Glimpse attains frontier-level performance at the level of (and sometimes exceeding) GPT-6 Astra and Claude Opus 5.5 while being a substantially smaller and cheaper model.”
Three Implications
TEAMS BUYING VISUAL AI Elorian lists Glimpse at $0.60 in and $3.00 out per million tokens, rates it says are 94% below GPT-6 Astra’s, with API access in a research preview.
FRONTIER LABS On Elorian’s own table Glimpse exceeds Astra on WGO Bench, CharXiv-R, UGC-VideoCap and FSC-147 and matches it on ERQA; Astra stays ahead on six other evaluations.
INVESTORS IN ELORIAN TechCrunch reports a $55 million seed round at a $300 million valuation co-led by Striker Venture Partners, Menlo Ventures and Altimeter Capital, with participation from Nvidia and Jeff Dean.
The framework opens: “Every technology stack has layers. Some are scarce, some are abundant.”
As we read it, Glimpse sits on that split: Elorian says it post-trained an open weights base model and sells the visual reasoning on top through a paid API. The framework predates the launch and does not discuss Elorian.
From Elorian’s Table 1 (16 October 2026): Glimpse’s scores minus GPT-6 Astra’s on the four evaluations where Elorian says Glimpse exceeds Astra. The point gaps are our arithmetic; Astra is ahead on six other evaluations in the same table.
What Is Not Established
The benchmark results are Elorian’s own evaluations; the texts we read do not report an independent test of Glimpse.
We read Elorian’s launch post and TechCrunch’s report on 16 October 2026. We did not contact Elorian, OpenAI or Nvidia, and we did not test Glimpse.
Elorian released Glimpse on 16 October 2026 at $0.60 per million input tokens and $3.00 per million output tokens, rates it says are 94% below GPT-6 Astra’s, and says the model exceeds Astra on four of its evaluations and matches it on a fifth.
Elorian says a small team built Glimpse on a few hundred GPUs by post-training an open weights base model. TechCrunch reports that Elorian raised a $55 million seed round at a $300 million valuation with participation from Nvidia.
A note on sourcing. We read in full Elorian’s launch post “Introducing Glimpse” (16 October 2026) and TechCrunch’s report of the same day on 16 October 2026; every quotation and figure comes from those texts, and the benchmark results are Elorian’s own. We did not contact Elorian, OpenAI or Nvidia and did not test Glimpse. The references to our earlier reporting link to our own pieces.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur, HackerNoon and Search Engine People. Find him on LinkedIn and Substack. See how we source.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
