Deepseek v4.1 flash prices cache-hit input at three dollars per billion tokens off-peak, while…
deepseek v4.1 flash prices cache-hit input at three dollars per billion tokens off-peak, while claude opus 5 charges five thousand dollars for the same billion, a 1,600x gap. the 552 billion parameter open-weight model carries a million token context and ships mit.
when the marginal cost of frontier-class reasoning hits fractions of a cent, the pricing conversation stops being about models.
Context
DeepSeek's pricing page, read on 4 October 2026, lists deepseek-flash as DeepSeek-V4.1-Flash with a 1M context. Per million tokens, cache-hit input is 0.003 dollars off-peak and 0.006 peak, cache-miss input is 0.15 and 0.30, and output is 0.60 and 1.20. Peak hours are 01:00-04:00 and 06:00-10:00 UTC, Monday to Friday. 0.003 dollars per million is 3 dollars per billion. The model card describes a multimodal mixture-of-experts model with 552B backbone parameters and up to one million tokens of context. Anthropic's pricing page lists Claude Opus 5 base input at 5 dollars per million tokens, 5,000 per billion, with cache hits at 0.50 per million and output at 25.
The note sets DeepSeek's cache-hit price against Opus 5's base, uncached input price. Cache hit against cache hit is 0.50 against 0.003 per million, about 167 times. Base input against DeepSeek cache hit is 5 against 0.003, about 1,667 times, so the note's 1,600x is a coarse rounding of that. Against DeepSeek's cache-miss price off-peak, Opus 5 base input is about 33 times. The off-peak qualifier applies to DeepSeek only. Both pages are current snapshots, not launch prices; a secondary snippet dates the price to 10 September 2026. The MIT license was not located in the model card text read, so it is not asserted here. The line that the pricing conversation stops being about models is the author's opinion.
Related work
- Earlier note on V4.1 Flash ↗Same model, different price tiers.
Watch next
- A dated DeepSeek release note and any Anthropic price change.
Sources
- DeepSeek API pricingapi-docs.deepseek.com
- DeepSeek-V4.1-Flash model cardhuggingface.co
- Anthropic pricingdocs.anthropic.com
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 18 September 2026 at 23:17 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 →