Elorian
An AI startup focused on vision-native reasoning architectures, which advocates using images directly as the medium of reasoning rather than translating images into text.
Core Facts
Timeline (last 90 days)
The Elorian team released a benchmark called "Decoy Counting" to evaluate the counting ability of multimodal large models when distracting visual elements are present.
Elorian is a new team based in Palo Alto, and founder Andrew Dai was formerly CEO of Google DeepMind.
Glimpse is Elorian's first model, released as a research preview with both an API and a chat interface.
Elorian released its first model, Glimpse, on October 16, positioned as a research preview and focused on visual thinking.
Glimpse is priced on Elorian's own platform at $600,000 per million input tokens and $3,000,000 per million output tokens.
Elorian claims that Glimpse's pricing is 94% cheaper than GPT-6 Astra's $10/$50.
The 12-row benchmark table released by Elorian shows that Glimpse won 4 rows, tied 1, and lost 6 against GPT-6 Astra, with Astra having no score in 1 additional row.
Elorian is a startup based in Palo Alto, and according to TechCrunch, it raised a $55 million seed round with a $300 million valuation before it had a product, with investors including Nvidia and Google's Jeff Dean.
Elorian CEO Andrew Dai was formerly a researcher at Google DeepMind.
Glimpse is built on an unnamed open-weight base model and post-trained by a small team with only a few hundred GPUs.
11 more timeline events
All Facts (20)
During its thinking process, Glimpse repeatedly returns to the original image, and can crop regions, extract frames from video, and run code to measure, which Elorian calls a visual scratchpad.
65%UnverifiedElorian is a new team based in Palo Alto, and founder Andrew Dai was formerly CEO of Google DeepMind.
50%UnverifiedGlimpse is built on an unnamed open-weight base model and post-trained by a small team with only a few hundred GPUs.
50%UnverifiedGlimpse is priced on Elorian's own platform at $600,000 per million input tokens and $3,000,000 per million output tokens.
50%UnverifiedElorian claims that Glimpse's pricing is 94% cheaper than GPT-6 Astra's $10/$50.
50%UnverifiedThe 12-row benchmark table released by Elorian shows that Glimpse won 4 rows, tied 1, and lost 6 against GPT-6 Astra, with Astra having no score in 1 additional row.
50%UnverifiedElorian is a startup based in Palo Alto, and according to TechCrunch, it raised a $55 million seed round with a $300 million valuation before it had a product, with investors including Nvidia and Google's Jeff Dean.
50%UnverifiedElorian CEO Andrew Dai was formerly a researcher at Google DeepMind.
50%UnverifiedDevelopers can call Glimpse through the AI SDK or the Chat Completions interface using the model name elorian/glimpse.
50%UnverifiedAs of the time the article was published, the specific contents of Elorian's release still could not be confirmed, and the original information was extremely limited.
50%UnverifiedElorian, as a team name, has not previously had extensive public coverage, and may be a relatively emerging team or project codename.
50%UnverifiedThe amount of information Elorian has currently released is limited, and key indicators such as the model's technical details, training data composition and inference efficiency have yet to be disclosed.
50%UnverifiedThe evaluation data released by Elorian has not yet been independently reproduced by a third party.
50%UnverifiedAccording to evaluation data released by Elorian, Glimpse surpassed GPT-6 Astra on both scientific chart reasoning and short video description tasks.
50%UnverifiedGlimpse adopts a 'vision-native reasoning' architecture, in which the intermediate representations during reasoning remain in visual form rather than being translated into text tokens early on.
50%UnverifiedThe Elorian team released a benchmark called "Decoy Counting" to evaluate the counting ability of multimodal large models when distracting visual elements are present.
50%UnverifiedGlimpse is Elorian's first model, released as a research preview with both an API and a chat interface.
50%UnverifiedElorian released its first model, Glimpse, on October 16, positioned as a research preview and focused on visual thinking.
50%UnverifiedElorian's Glimpse visual reasoning model has been integrated into AI Gateway, accepts image and video inputs, and can track objects through motion, judge spatial relationships, and link actions to physical consequences.
50%UnverifiedA social platform tweet mentioned the latest results from AndrewDai and the Elorian team, and the tweet read 'Have a glimpse at this! Nice work, @AndrewDai and Elorian folks!'
50%