Tag
All content about DeepSeek, organized for fast scanning.
DeepSeek has recently launched several updates, including the DeepSeek Harness v0.2, which introduces desktop applications for macOS and Windows, enhancing document workflows and task management. The company also released DeepSeek-V4-Pro, featuring improved agent capabilities and a new API pricing structure, while the V4 Flash is now in public beta with upgraded functionalities. Additionally, a closed beta for DeepSeek Harness invites open-source developers to contribute, highlighting the platform's focus on community engagement and innovation.
DeepSeek has just rolled out DeepSeek Harness v0.2, bringing packaged desktop apps for macOS and Windows. The update expands document and dev workflows, adds recurring task scheduling, and introduces a built-in plugin manager and creator mode.
DeepSeek has just rolled out DeepSeek-V4-Pro, bringing adjustable reasoning levels, stronger agent workflows, and native OpenAI Responses API support. It also introduces peak/off-peak API pricing starting Aug. 16, 2026, sparking debate over costs versus gains.
DeepSeek Harness has entered a closed beta, with invites aimed at developers building open-source projects tied to Agent Harness. Jiayuan Zhang says interested builders should reply with a GitHub ID and representative project details.
DeepSeek has just rolled out V4 Flash in public beta, with the official API now live and upgraded agent capabilities. Company-posted benchmarks show gains over prior DeepSeek preview builds, though several results still trail Opus-4.8.
A recent benchmark write-up by Darko at Kilo Blog tests DeepSeek V4 Pro and Flash against Claude Opus 4.7 and Kimi K2.6 using a heavier FlowGraph workflow. It digs into where each model shines, stumbles, and how pricing shifts the value equation.
DeepSeek has just rolled out a 75% discount on its DeepSeek-V4-Pro API through May 5, 2026. The company also teased agent-focused integration updates, including Claude Code support for a 1M-context model variant.
DeepSeek has just rolled out DeepSeek-V4 Preview, open-sourcing the model and making 1M context the default across its services. The launch includes V4-Pro and V4-Flash, plus new sparse attention and compression claims, fresh API pricing, and a planned 2026 retirement for older models.