DeepSeek Releases V4 Pro With Higher Benchmarks, Open-Source Tooling, and Upcoming Price Increases

Chinese AI company DeepSeek released DeepSeek-V4-Pro-0813 on its web interface, mobile app, and API, claiming it matches rivals like Moonshot AI's Kimi K3 at lower cost. The model scored 62.7 on the DeepSWE benchmark and 87.9 on Terminal Bench 2.1, but 53 on Artificial Analysis's Intelligence Index, below Kimi K3's 60 and Anthropic's Claude Opus 5's 63. DeepSeek also open-sourced DeepSeek Harness (dsh) v0.1 under the MIT license and announced API price increases of 50% to 1,100%, with peak-hour output at $3.96 per million tokens starting August 16, 2026.

DeepSeek Releases V4 Pro With Higher Benchmarks, Open-Source Tooling, and Upcoming Price Increases Chinese AI company DeepSeek https://www.ghacks.net/2026/04/26/deepseek-releases-v4-models-with-9-5x-lower-memory-requirements-and-huawei-ascend-support/ has launched the full version of its DeepSeek-V4-Pro-0813 https://api-docs.deepseek.com/news/news260813/ model on its web interface, mobile app, and API. The model targets autonomous AI agent tasks and software engineering. DeepSeek claims it matches the capabilities of models such as Moonshot AI's Kimi K3, but at a lower cost. The company also released an open-source developer tool and announced changes to its API billing, with price increases ranging from 50% to 1,100%. This release is a significant update compared to the preview version that was introduced in April. DeepSeek V4 Pro Improves Benchmarks While Staying Cheaper Than Rivals DeepSeek reports that V4 Pro scored 62.7 on the DeepSWE benchmark for real-world software engineering, a notable increase from 12.8 in earlier testing. The model also achieved a score of 87.9 on Terminal Bench 2.1. The independent benchmarking firm Artificial Analysis gave the reasoning version of V4 Pro a score of 53 on its Intelligence Index. This is lower than Moonshot's Kimi K3 at 60 and Anthropic's Claude Opus 5 at 63. However, DeepSeek's pricing remains well below both competitors. On third-party platforms such as OpenRouter, V4 Pro is priced at $0.435 per million input tokens and $0.87 per million output tokens. In comparison, Moonshot's Kimi K3 costs between $14 and $15 per million output tokens, and US competitors like Anthropic's Claude Fable 5 can reach $50 per million. DeepSeek Harness, API Billing Changes, and Availability DeepSeek also released DeepSeek Harness dsh v0.1 under the open-source MIT license, positioning it as an alternative to Anthropic's Claude Code. Built on a modular architecture in which "everything is a plugin," the tool lets developers orchestrate autonomous coding agents across local file systems and development. DeepSeek is also changing its API billing. The company will move from flat rates to peak and off-peak pricing, with API costs increasing by 50% to 1,100% depending on the model and usage time. During peak hours, V4 Pro output will cost $3.96 per million tokens. This is a significant increase from current rates, but still less than many Western alternatives.rFor startups and enterprise teams running large-scale AI workloads, these changes have practical implications: - Compare per-task cost against capability, since a model slightly weaker on some benchmarks but far cheaper to run can still be the better business decision at scale. - Factor in the August 16 pricing change, which moves V4 Pro into a premium tier while keeping it below many US competitors. - Consider the DeepSeek Harness tool if evaluating alternatives to Claude Code for orchestrating coding agents. - Weigh peak versus off-peak usage hours, since the new billing structure ties cost to when workloads run. DeepSeek-V4-Pro-0813 is now available through DeepSeek's web interface, mobile app, and API. The new peak and off-peak pricing will start on August 16, 2026. According to eWeek, DeepSeek is testing if moving V4 Pro into a premium tier can support a sustainable business, instead of relying on low pricing. The long-term pricing plan has not been confirmed.