Roupen Odabashian, Oncologist at Abbotsford Regional Hospital and Cancer Centre, Founder of MeDucation AI, Podcast Host at OncoDaily, shared on LinkedIn:
”Claude Opus 5 launched at $5 and $25 per million tokens, roughly half the price of the model it replaced, while topping the Artificial Analysis Intelligence Index at 61. A week later DeepSeek V4-Flash landed at $0.14 and $0.28.
Everyone is watching the capability line. The price line is the one that changes what gets built in medicine.
A rough sense of scale: a full oncology chart, notes, pathology, imaging reports, prior lines of therapy, is a few hundred thousand tokens. Two years ago running frontier reasoning over one patient’s complete record was a line item you had to defend. It is now closer to the cost of the coffee you drink while reading it.
That does not make the output clinically valid. Validation is still the hard part and no pricing change touches it.
But it does mean the thing that was blocking a lot of clinical AI was never the model. It was that nobody could justify running the good model over every patient, so we ran the cheap model over some patients and called it a pilot.
When inference gets close to free, the question stops being
‘can we afford to check every chart’
and becomes
‘who is accountable for what the check says.’ ”
You can also read ‘3rd Symposium ‘Artificial Intelligence in Public Health Research’ ‘
