Skip to main content

Tencent Slashes AI Model Prices by Up to 97.5%

Tencent's Bold Move in AI Pricing

In a move that's set to shake up the AI development landscape, Tencent Cloud announced sweeping price reductions for its DeepSeek-V4 series models. Starting June 3, 2026 at midnight, developers will see costs plummet by up to 97.5% for certain services.

The Numbers Speak Volumes

The DeepSeek-V4-Pro model leads the price cuts with both input and output costs slashed by 75%. Processing a thousand tokens now costs just 0.003 yuan for input and 0.006 yuan for output. But the real jaw-dropper? The cache hit price for V4-Pro has been reduced to 0.000025 yuan per thousand tokens - a staggering 97.5% reduction from previous rates.

Not to be outdone, the DeepSeek-V4-Flash model also enjoys a 90% price cut for cache hits, matching the same ultra-low rate of 0.000025 yuan per thousand tokens.

Image

Why This Matters

Launched just this April, the DeepSeek-V4 series has quickly become a favorite among AI developers. Its 1.6 trillion parameters and mixture of experts (MoE) architecture make it one of the most powerful models available, with native support for contexts up to one million tokens.

"This isn't just about making AI cheaper - it's about making cutting-edge technology accessible," explains industry analyst Li Wei. "At these prices, we'll see startups and smaller businesses experimenting with AI in ways that were previously unimaginable."

A Pattern of Affordability

Tencent had already signaled its commitment to affordable AI earlier in May, when it made temporary price reductions for the V4-Pro API permanent. This latest move suggests the tech giant is serious about dominating the AI-as-a-service market through aggressive pricing strategies.

Developers are already celebrating the news. "This changes everything," shared Zhang Ming, CTO of a Shanghai-based AI startup. "Our monthly costs for model access just went from being our biggest expense to almost negligible."

Key Points

  • Massive price cuts: Up to 97.5% reduction on DeepSeek-V4 services
  • New rates effective June 3, 2026
  • V4-Pro input/output costs down 75%
  • Cache hit prices slashed to 0.000025 yuan per thousand tokens
  • 1.6 trillion-parameter models now dramatically more affordable