Model update
DeepSeek-V4-Pro's 75% discount is now permanent.
DeepSeek has confirmed that the reduced price for DeepSeek-V4-Pro is here to stay for good. It's an aggressive move: the official cost is now locked in at 1/4 of the original price.
On quality, it doesn't match GPT-5.5 or Claude Sonnet for general use. But for certain tasks — summaries, classification, extraction, drafts, or high-volume workflows — it can be a very useful model.
DeepSeek is already available in every hablo agent.
The practical advice is to use it when cost is the priority and the task doesn't demand the level of precision, reasoning, or reliability of the flagship models.
DeepSeek-V4-Pro's new permanent pricing
| Model | Input cache hit | Input cache miss | Output |
|---|---|---|---|
| DeepSeek-V4-Pro | $0.003625 $0.0145 |
$0.435 $1.74 |
$0.87 $3.48 |
API price announced as permanent following the extended offer: 1/4 of the original price.