Newsletter Subscribe
Enter your email address below and subscribe to our newsletter
[forminator_form id="25163"]

venturebeatventurebeatventurebeatDeepSeek on Thursday released an open-source agent framework designed to challenge Anthropic's Claude Code and OpenAI's Codex, marking the Chinese AI lab's most direct move yet into the software layer that controls how AI models execute complex coding tasks.
DeepSeek Harness v0.1, available under the MIT license, is built on a system called Cordis where "everything is a plugin" — models, tools, sandboxes, filesystems, orchestration and user interfaces can all be swapped out by developers. The framework can inspect repositories, edit files, execute shell commands, maintain plans, invoke skills and delegate work to subagents.venturebeat
The release arrived alongside DeepSeek-V4-Pro-0813, an updated version of the company's 1.6-trillion-parameter flagship model with what the company describes as "significantly enhanced Agent capabilities". V4-Pro is now available across DeepSeek's web interface, mobile app and API, with native support for the OpenAI Responses API and integration with Codex.venturebeat
DeepSeek's GitHub repository had already attracted roughly 27,500 stars and 2,000 forks as of Thursday. The project remains in developer preview, with DeepSeek warning that breaking changes should be expected.venturebeat
The technical launch follows a broader organizational push. DeepSeek created an official WeChat account for its "DeepSeek Harness Team" and posted job listings in Beijing for a Product Manager and R&D Engineer, according to Bloomberg. The team began assembling in May 2026. A recruitment notice states the team's goal is to develop DeepSeek models into "frontier intelligent-agent products".tradingview+2
Unlike Claude Code and Codex, which offer hosted background agents, GitHub-native PR workflows and broad IDE integrations, Harness emphasizes radical modularity — the model itself is just another replaceable plugin rather than the center of a vertically integrated stack.venturebeat
DeepSeek is simultaneously raising its API prices. Beginning August 16 at 16:00 UTC, the company will replace flat pricing with peak and off-peak rates. V4-Pro output pricing will rise from $0.87 per million tokens today to $1.98 off-peak and $3.96 during peak hours. Even off-peak rates represent more than double current levels.venturebeat
The price increases reverse a trajectory that saw DeepSeek cut V4-Pro rates by 75% between April and late May. For organizations already running DeepSeek in production, workload scheduling and the option to run open weights on their own infrastructure become more important considerations. DeepSeek is now pursuing two potentially conflicting advantages: making its agent stack more open while making its hosted API considerably more expensive.venturebeat