No ad buys: DeepSeek uses its technical report as the launch asset

Brand advertising persuades developers badly. They do not believe the launch film; they believe benchmarks, weights and invoices. A model from China also enters with a pre-set distance: overseas communities start sceptical.

There are really only two levers strong enough to close that distance quickly — something developers can verify themselves, and a price nobody can ignore.

DeepSeek used both, folded into one move: publish everything at release. R1 shipped with the full technical paper and model weights, alongside API pricing cut to roughly one-twentieth of comparable tiers.

Distribution ran through verification venues rather than ad buys: GitHub, Hugging Face and Discord carried verification and support, while X amplification came from tech opinion leaders endorsing it unprompted. The report judges this organic spread more effective than paid influencer work.

Replace content marketing with the technical report — model card, paper and weights placed where they are easiest to reproduce.

Run Discord as front-line support; the community grew alongside attention and absorbed both support and feedback.

Use pricing as an entry ticket rather than a discount frame, so cost itself becomes a talkable point for overseas startups adopting it as infrastructure.

Checkable: the article explicitly states no paid media and no offline advertising, API pricing of about one-twentieth of comparable OpenAI performance tiers, the paper and weights shipped together with R1, and Discord membership "passing hundreds of thousands" without an exact figure.

Not checkable: no quantified X metrics, no GitHub star count given, and "more effective than paid influencer marketing" is the author’s judgement rather than a verifiable comparison. Both are flagged here rather than converted into numbers.

For developer-facing products, verification cost sets the speed of spread. Releasing weights outsources the persuading to the user.

Price only becomes shareable when it can be restated. "One-twentieth" is a sentence anyone can repeat; "leading cost-performance" is not.

Independent observation, not confirmed with DeepSeek; figures follow the source’s description of early 2025.

Discord membership and star counts lack precise definitions and timestamps. No growth curve is estimated and no paid conversion inferred.