A permanent 75% discount on the high-performance Chinese AI 'DeepSeek-V4-Pro' is being offered, and attention is also being drawn to 'Reasonix,' a coding agent specifically designed for DeepSeek.



DeepSeek, a Chinese AI company, announced the DeepSeek-V4 series in April 2026 and offered a limited-time 75% discount on API fees. Now that it has been announced that this discount will be permanent, the DeepSeek-V4-Pro is attracting attention as a cost-effective model.

Models & Pricing | DeepSeek API Docs

https://api-docs.deepseek.com/quick_start/pricing

DeepSeek announced 'DeepSeek-V4' on April 24, 2026. DeepSeek-V4 comes in two versions: 'DeepSeek-V4-Flash' with a total of 284 billion parameters and 13 billion active parameters, and 'DeepSeek-V4-Pro' with a total of 1.6 trillion parameters and 49 billion active parameters. Both are distributed free of charge as open models, and can also be run via DeepSeek's paid API.

The 'DeepSeek-V4' has finally arrived, an open-top model with performance exceeding that of the Claude Opus 4.6 - GIGAZINE



DeepSeek is promoting DeepSeek-V4-Pro as outperforming closed models such as 'Claude-Opus-4.6-Max,' 'GPT-5.4-xHigh,' and 'Gemini-3.1-Pro-High.' The high performance of DeepSeek-V4-Pro has also been confirmed by third-party tests, with the Center for Artificial Intelligence Standardization and Innovation (CAISI), the AI evaluation body of the National Institute of Standards and Technology (NIST), reporting that DeepSeek-V4-Pro performs comparably to GPT-5. CAISI also assessed DeepSeek-V4-Pro as being 'about eight months behind major American AI models,' which implies that DeepSeek, a Chinese company with limited access to high-performance NVIDIA GPUs, is able to keep up with American companies, albeit only eight months behind.

'DeepSeek V4 Pro is about eight months behind major US AI models, but is currently the most high-performing Chinese-made AI model,' reports CAISI, the US government's AI risk management agency - GIGAZINE



DeepSeek-V4-Pro was launched and offered a limited-time 75% discount on API fees until May 5, 2026. The API fees for DeepSeek-V4-Pro per 1 million tokens were normally $1.74 (approximately 276.6 yen) for input, $0.0145 (approximately 2.3 yen) for cached input, and $3.48 (approximately 533.1 yen) for output. However, with the 75% discount, the fees became $0.435 (69.1 yen) for input, $0.003625 (0.6 yen) for cached input, and $0.87 (138.3 yen) for output.




Subsequently, on April 29, 2026, it was decided to extend the discount period until May 31, 2026.




Then, on May 23, 2026, it was announced that the 75% discount would be made permanent.




The following is a price-performance ratio graph of AI published by Artificial Analysis, which performs AI performance analysis. The horizontal axis represents price and the vertical axis represents performance, and it can be seen that the price of DeepSeek-V4-Pro is very low compared to AI with equivalent performance.



The permanent discount on DeepSeek-V4-Pro has become a major topic of discussion online, and the news sharing site Hacker News is also highlighting Reasonix , a third-party coding agent specifically for DeepSeek.

Reasonix — DeepSeek-native AI coding agent for your terminal
https://esengine.github.io/DeepSeek-Reasonix/



Related Posts:

in AI, Posted by log1o_hf