DeepSeek price slash fuels competition, hits Zhipu, Minimax

Invezz | April 27, 2026 at 07:46 AM UTC
Bearish 84% Confidence Unanimous Agreement
Read Original Article

Key Points

  • DeepSeek introduced a two-tier strategy with V4-Pro (high performance) and V4-Flash (budget-friendly), targeting complex AI agent tasks while claiming to outperform open-source rivals and trail only Google's Gemini-Pro-3.1
  • The V4 model is adapted to run on Huawei chips, representing a strategic test of China's ability to scale advanced AI using domestic infrastructure amid U.S. export controls on Nvidia hardware
  • Competitors face margin pressure as DeepSeek's price cuts compress API pricing across the sector, with rivals like Tencent also launching new models (Hy3) to defend market position

AI Summary

Summary

Key Development:

Chinese AI startup DeepSeek launched its fourth-generation model (V4) with aggressive pricing cuts, offering developers a 75% discount on its DeepSeek-V4-Pro model until May 5, 2026, and reducing input cache hit prices across its API lineup to one-tenth of original costs.

Market Impact:

The announcement triggered significant sell-offs among competitors:

  • Zhipu AI (Knowledge Atlas Tech, HK:2513) fell 3.4%, following a 9% decline on Friday
  • MiniMax Group (HK:0100) dropped 10%, extending a prior 9% loss
  • Analysts recommend selling both stocks, citing margin compression and limited differentiation in a price-war environment

Product Strategy:

DeepSeek introduced a two-tier approach with a Pro model (outperforming open-source competitors except Google's Gemini-Pro-3.1) and a lighter Flash variant. The V4 models target complex AI agent tasks and represent a major upgrade from the December 2024 V3 launch.

Strategic Significance:

The V4 model is adapted for Huawei's chip technology, signaling China's push for domestic AI infrastructure alternatives amid US export controls. This positions the launch as both a commercial and geopolitical development.

Competitive Context:

The pricing strategy intensifies competition in China's AI sector, where API costs have become a critical battleground. Tencent recently launched its Hy3 model under former OpenAI leadership, adding pressure. DeepSeek's move raises broader concerns about an AI industry price war affecting profitability across the sector.

Outlook:

The preview rollout aims to gather real-world feedback before full release, with no detailed performance benchmarks or final pricing disclosed yet.

Model Analysis Breakdown

Model Sentiment Confidence
GPT-5-mini Bearish 80%
Claude 4.5 Haiku Bearish 82%
Gemini 2.5 Flash Bearish 90%
Consensus Bearish 84%