LUNAとは

On August 14, DeepSeek's product line underwent a series of updates, officially launching DeepSeek-V4-Pro and its first Agent product, DeepSeek Harness. At the same time, DeepSeek-V4-Pro-0813 was officially launched on Hugging Face, with weights available for download, continuing under the MIT license. The official release also provided vLLM, SGLang, and local operation instructions, allowing developers to deploy this version independently. In terms of product capabilities, the previously leaked evaluation results of the DeepSeek-V4-Pro Agent have been confirmed. The official version of DeepSeek-V4 has also seen a price adjustment, with the peak output price for Pro increasing by 4.5 times compared to before. Analysts have noted that after the price increase, DeepSeek's peak Flash price has surpassed that of GPT-5.6 Luna, indicating that its low-price advantage is being re-evaluated. The Agent direction is the core highlight of this update. DeepSeek Harness has been released to npm, positioned as DeepSeek's first Agent product, supporting project management, long tasks, multiple Agents, Skills, online search, and context management, while allowing users to create different Agent presets. Tasks such as coding, research, and writing can be configured with different tools, prompts, and workflows. Architecturally, DeepSeek Harness adopts 'Everything is a plugin' as its core design, allowing models, tools, prompts, storage, context, and even interfaces to be replaced or combined through plugins. Its underlying framework, Cordis, has also publicly released its design paper, breaking down model adapters, tool systems, Session Logs, and Agent Loops into replaceable components, supporting runtime loading, unloading, and combination. Market interest has rapidly increased. Approximately an hour and a half after the open-sourcing of DeepSeek Harness, GitHub Stars surpassed 22,000, exceeding the previous public records of xAI Grok-1 and DeepSeek-R1, which both broke 20,000 Stars. Industry focus has shifted from single model capabilities to Agent runtime, plugin systems, and the engineering foundation required for recursive self-improvement.

According to monitoring by Beating, OpenAI has reduced the price of GPT-5.6 Luna by 80% and GPT-5.6 Terra by 20%. The LUNA API is priced at $0.20 per million input tokens and $1.20 per million output tokens. SOL pricing, ChatGPT and Codex subscription prices, and quota budgets remain unchanged; Terra and Luna consume fewer credits. The GPT-5.6 SOL fast mode has replaced priority processing in the API. Click the original link below to join the Beating · Feishu AI news channel for 24/7 monitoring of global AI highlights and news.

On July 31, OpenAI reduced the price of GPT-5.6 Luna by 80% and GPT-5.6 Terra by 20%. The Luna API is priced at $0.20 per million input tokens and $1.20 per million output tokens; the Terra API is priced at $2 per million input tokens and $12 per million output tokens; SOL pricing, ChatGPT and Codex subscription prices, and quota budgets remain unchanged; Terra and Luna consume fewer credits.