No AI release moved the market like DeepSeek’s in January 2025: frontier-adjacent results at a fraction of the assumed training cost, weights published, licences permissive. Whatever else changes, it settled one argument permanently: the open-weight route is a serious business option, not a compromise, and it forced every price in the industry toward honesty.

What you are choosing between

Three very different DeepSeeks hide under one name:

  • The open-weight models. The family that matters most for firms: downloadable, MIT-style licensed, served by many third-party hosts, including through the routing this estate itself uses. Run this way, privacy equals your hosting choices.
  • The direct API. Priced aggressively, processing with the provider under its terms; judge it with the standard checklist like any processor.
  • The consumer app. Briefly topped the download charts, and belongs on the same approved-or-not line as every consumer AI app in your tools policy.

The business case, honestly

Capability per pound is the whole pitch, and it is real: for extraction, classification, drafting and reasoning-heavy backroom work, DeepSeek-class models through a chosen host deliver most of the frontier at a fraction of the cost, which is precisely the open-weight thesis.

The considerations on the other side are equally real and unembarrassing to name:

  • Provenance matters to some boards and some clients, and pretending otherwise helps nobody; put the question to your risk owner and record the answer.
  • Polish is thinner than the frontier products: tooling, enterprise contracts, support depth.
  • Adoption route for regulated firms is via hosts and terms you can paper, not the consumer surface. This site’s benchmark method applies unchanged: five of your tasks, through a host, judged on output.

Where I fit in

DeepSeek-class models sit in my fallback chains where their economics win, on hosts chosen deliberately, and the same option is open to client builds where the workload and the risk appetite agree. The decision is the open-weight decision with one extra question attached, and an Automation Audit answers both with your numbers rather than the discourse’s.