DeepSeek launches its V4 Pro model into full release, boasting significant benchmark improvements, new open-source tools, and a radical pricing strategy aimed at making autonomous AI more accessible for software engineering applications.
DeepSeek has moved its V4 Pro model from preview to full release across its web interface, mobile app and API, marking a substantial update to a system aimed at autonomous agents and software engineering. The company says the latest version is intended to compete with stronger proprietary models while remaining cheaper to run, a positioning that has become central to its product strategy. The release also broadens access to the model family through both direct and third-party channels.
Benchmark claims suggest a sizeable step forward. DeepSeek says V4 Pro scored 62.7 on the DeepSWE test for real-world software engineering, up from 12.8 in earlier testing, and 87.9 on Terminal Bench 2.1. Artificial Analysis, an independent benchmarking firm, gave the reasoning variant an Intelligence Index score of 53, below Moonshot AI’s Kimi K3 at 60 and Anthropic’s Claude Opus 5 at 63, but DeepSeek’s pricing remains far lower than those rivals, according to published rates on OpenRouter.
That cost gap is part of the model’s appeal. On OpenRouter, V4 Pro is listed at $0.435 per million input tokens and $0.87 per million output tokens, while some competing systems from Moonshot and Anthropic are priced many times higher. DeepSeek appears to be testing whether a model that is not always top-ranked on benchmarks can still win business by offering enough capability at a much lower inference cost, especially for teams running large workloads.
Alongside the model launch, DeepSeek has released DeepSeek Harness, or dsh v0.1, under the MIT licence. The open-source tool is presented as a rival to Anthropic’s Claude Code and uses a modular design in which, as the company puts it, “everything is a plugin”. That structure is meant to help developers coordinate autonomous coding agents across local file systems and development environments.
The company is also changing its API billing model. Instead of flat pricing, DeepSeek will move to peak and off-peak rates, with increases ranging from 50% to 1,100% depending on the model and the time of use. During peak hours, V4 Pro output will cost $3.96 per million tokens, a sharp increase from earlier rates but still below many Western alternatives. DeepSeek-V4-Pro-0813 is now available through its web interface, mobile app and API, with the new billing structure taking effect on 16 August 2026.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
