Product / UX / UI & API

DeepSeek Chat

Hosted consumer chat product from DeepSeek, at chat.deepseek.com and in iOS and Android apps, fronting the company's own model family with long-context handling, file reading and agent capabilities. It runs as a free hosted service, distinct from the model weights of the same name.

The underlying DeepSeek models are scored separately under base_pretrained; this record covers the chat product. Verified 2026-08-13 via the DeepSeek product site.

Openness

1 high confidence

1.0

client: proprietary(no public source for chat.deepseek.com app)
backend: hosted-SaaS
note: underlying models open(MIT) but the chat application/service is closed
source: closed(hosted chat service at chat.deepseek.com

The DeepSeek chat surface is a proprietary hosted service, and the openness of the V3/V4 weights it serves is scored separately at the model layer. It is the closed counterpart to the open chat interfaces in this category.

https://www.deepseek.com/en/ recorded 2026-08-13
Free hosted chat at chat.deepseek.com serving DeepSeek-V4 Preview; no source release for the chat app itself
https://www.deepseek.com/en recorded 2026-08-13
The site offers exactly two ways to use DeepSeek - 'Free access to DeepSeek' pointing at chat.deepseek.com, and 'Build with the latest DeepSeek models' pointing at the hosted API platform with its own pricing. The Research section links GitHub repositories (DeepSeek R1, V3, Coder V2 and others), and every one of them is a model-weights or research/inference release; none is the chat application. There is no self-hostable implementation of the chat surface being scored here, which is what the ladder calls a closed source.

Adoption

5 medium confidence

5.0

130M monthly active users as of May 2026, third among Chinese AI apps, per QuestMobile's "2026 First-Half AI Application Market Development Insight Report". That is far above the >10M users threshold, level 5 on the active-user scale. Confidence is medium because QuestMobile's report is read through press coverage rather than the report itself, and it measures the China market, where DeepSeek's usage is concentrated but not confined. The count belongs to the chat surface rather than to any particular model release.

https://finance.biggo.com/news/372811e2-28f1-49c7-9f53-fa56879cde3b recorded 2026-08-13
QuestMobile 2026 H1 report, as of May 2026: "DeepSeek ranked third with 130 million MAUs". Doubao leads at 382M.
https://play.google.com/store/apps/details?id=com.deepseek.chat&hl=en_US recorded 2026-06-04
Official DeepSeek AI Assistant Android app listing (consumer chat surface, mass-market distribution)

Capability

3 medium confidence

3.0

A capable single-vendor consumer chat interface over a strong underlying model, but narrow on the dimensions this category measures - one provider, limited enterprise, RAG and plugin breadth - compared with frontier multi-model chat interfaces, so not a category frontier-definer. The current family is DeepSeek-V4-Pro, reached through web, app and API, with enhanced agent capabilities plus a Responses API and Codex integration, and still no provider menu, enterprise multi-user tier or grounding over corporate data.

https://www.deepseek.com/en/ recorded 2026-08-13
Chat product features: file reading, long-context, V4 agent capabilities, single-vendor model

Verified 2026-08-13