AI news · October 17, 2026
Elorian’s Glimpse treats images as a workspace for reasoning
Elorian’s Glimpse offers API access at $0.60 input/$3 output per million tokens and claims strong chart and video results; the real question is repeatable accuracy on your own media.
- USD per million input tokens
- 0.60
- USD per million output tokens
- 3.00
- GPUs used for training, company reported
- few hundred
Elorian released Glimpse on October 16 as a research preview with API access; its chat interface is waitlist-only. The model is built to revisit visual evidence during a task: it can crop a region, pull selected video frames, and run code to measure or compare what it sees. TechCrunch reports a small team trained it on a few hundred GPUs and says the model targets robotics, science, media, and creative work.
Elorian says Glimpse costs $0.60 per million input tokens and $3 per million output tokens. Its own tables put it ahead of GPT-6 Astra on some chart, counting, video, and manipulation evaluations, but the comparisons are vendor-run and other tests still favor larger models. API access makes this one of the few stories here that can move straight into a controlled builder trial.
What you can do with it
Try Glimpse on a small set of charts, short clips, and object-counting tasks. Save the input, answer, tool calls, latency, and review result. Compare it with the model you already use; stop if the lower price does not survive retries and human correction.
Our take
This is a useful model design choice, not proof that visual agents are solved. Rechecking the source can reduce errors that begin at first glance, but it can also add calls and delay. The vendor table is a starting point; an accepted-output rate on your data is the decision metric.
Links Try Glimpse
Source: TechCrunch — Made With Models writes the brief; the reporting is theirs.