The context · January 2025
DeepSeek released R1 in January 2025, bringing another openly available reasoning model into the conversation about how organisations build and operate AI. [1]
The decision behind the announcement
For a business choosing an AI platform, access to model weights changes the range of possible deployments. It does not remove the need for infrastructure, specialist support or evaluation. A service that looks inexpensive per request can become expensive when the team has to operate it around the clock.
Compare the work, not the headline
Start with a task your team already understands: extracting fields from a document, searching a technical handbook or drafting a support response. Compare accepted answers, review time and response latency. Include failed requests and retries in the cost calculation. A fluent answer that needs rewriting is still unfinished work.
Keep the exit practical
Model independence is most useful when the application can change providers without being rebuilt. Keep business rules, reference data and evaluation examples outside a provider-specific prompt. Preserve a simple fallback. The question is not whether an open model wins every benchmark, but whether your organisation can make and revisit an informed choice.
Source & context
DeepSeek · R1 release, 20 January 2025This retrospective was written for the archive in September 2026. The linked primary source documents the announcement or event; the practical interpretation and proposed approach are Sansa’s editorial perspective. Public examples do not imply a client relationship. Product capabilities and guidance may have changed since the period discussed.
Another perspective · January 2025
A small, useful test of hosted and self-managed AI
Continue readingWorking through a similar question?
Talk it through with Sansa