The llm price war hit a floor nobody predicted this fast. by september, openai's gpt-4o mini fell…
the llm price war hit a floor nobody predicted this fast. by september, openai's gpt-4o mini fell to $0.10 per million input tokens from $0.60 in january, with anthropic's sonnet 4.5 dropping 80% to $0.75.
when inference costs this little, the margin is in everything except the tokens.
Context
Two vendor pages were read as currently published, and neither is a price history. OpenAI's GPT-4o mini page lists input at $0.15, cached input at $0.075 and output at $0.60 per million tokens. Anthropic's pricing page lists Claude Sonnet 4.5 at $3 per million input tokens and $15 per million output tokens. A forum post dated July 30, 2026 reports price cuts on different OpenAI models (GPT-5.6 variants), summarized second hand.
The note's claimed cuts were not located. The $0.60 that the note gives as a January input price is the current output price on OpenAI's page, which looks like an input and output mix-up, and no source shows $0.10 input. Sonnet 4.5 is shown at $3 and $15, and 80% off $3 would be $0.60 and not $0.75, so the note's own figures disagree. These are current rates, so they do not prove what the prices were in January or September, and a cut at some other date is not excluded. A floor needs a matched basis, and the note mixes two vendors and input-only figures with no dated source. The only dated cut found concerns other models.
Watch next
- A dated price history or vendor changelog for either claimed cut.
Sources
- OpenAI docs: GPT-4o minidevelopers.openai.com
- Anthropic docs: pricingdocs.anthropic.com
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 22 September 2026 at 01:47 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →