TideAPI
All Posts
Apr 27, 20263 min readModel Launch

DeepSeek V4 Models Now Available: V4-Pro and V4-Flash

DeepSeek's latest generation is live on TideAPI — the flagship V4-Pro and the ultra-affordable V4-Flash, offering up to 96% savings vs GPT-5.4.

We're excited to announce that DeepSeek's brand-new V4 family is now live on the TideAPI platform. This includes two models — DeepSeek-V4-Pro, a flagship-grade general model, and DeepSeek-V4-Flash, an ultra-affordable option that delivers exceptional value at a fraction of the cost of leading Western competitors.

DeepSeek-V4-Pro — Flagship Performance

V4-Pro is DeepSeek's most capable model to date. With a 1M token context window and up to 384K output length, combined with improved reasoning, instruction following, and multilingual capabilities, it competes directly with top-tier models while remaining significantly more cost-effective.

Input

$3.67 / 1M tokens

Output

$7.33 / 1M tokens

Context

1M tokens

Max Output

384K tokens

DeepSeek-V4-Flash — Unbeatable Value

V4-Flash is designed for high-throughput, cost-sensitive workloads. With the same 1M context window and 384K max output as V4-Pro, at just $0.31 per million input tokens, it delivers 88–96% savings compared to GPT-5.4 — making it ideal for customer service, content generation, and batch processing tasks.

Input

$0.31 / 1M tokens

Output

$0.61 / 1M tokens

Context

1M tokens

Max Output

384K tokens

Savings vs GPT-5.4

Up to 96%

Previous Models Stay Available

DeepSeek-V3.2, V3.1, and R1-0528 remain fully available on the platform for existing integrations. We recommend evaluating V4-Flash as a drop-in replacement for V3.2 workloads — it offers better performance at a lower price point.

Getting Started

Both V4 models are available immediately through the TideAPI unified API. No changes to your integration are required — simply update the model parameter in your API calls.

  • Use model: "deepseek-v4-flash" for cost-optimized workloads
  • Use model: "deepseek-v4-pro" for tasks requiring maximum capability
  • Existing V3.x integrations continue to work without changes

Ready to get started?

Access all models through one unified API — no vendor lock-in, no complexity.