Updates·September 11, 2026, 09:41

New service promises cheap AI for developers using DeepSeek model

AI-generated from the sources listed.

The company Dolphin has launched an API based on DeepSeek's V4.1 Flash model and promises far lower prices than the major models. The claim is Dolphin's own marketing and has not been independently documented.

The company Dolphin, described as an AI infrastructure provider, has launched the Dolphin Inference API. The launch is dated October 11, 2026, and the service runs on the DeepSeek V4.1 Flash model.

According to Dolphin's own marketing, the service delivers coding and agent performance on par with Opus 5 at a price 97 percent lower than the leading models. No independent documentation has been presented to support the claim. It also is not clear what prices Dolphin charges.

The first 1,000 users to sign up get $30 in free credits. The service is aimed at developers who can only use API-based inference and not subscriptions. An API is a technical gateway that lets developers send tasks directly to an AI model from their own programs. Inference means the model is used to produce answers.

The model behind it

DeepSeek released V4.1-Flash on September 10, 2026. It is a so-called mixture-of-experts model with 552 billion parameters. That means only part of the model is in use at any given time: 8 billion active parameters for input and 16 billion for output.

According to DeepSeek, V4.1-Flash outperforms the company's earlier V4-Pro on performance, price, speed and total runtime. That is DeepSeek's own claim. According to DeepSeek, the model also requires a quarter of the HBM memory and an eighth of the SSD storage for KV cache compared with the previous generation. KV cache is a kind of working memory the model uses during a conversation. The model also has built-in image understanding.

What is planned

The model is available in DeepSeek's own API as deepseek-flash and as open source on Hugging Face. From September 14, 2026, requests to V4-Pro are to be redirected to V4.1-Flash. In DeepSeek's API, off-peak prices are 50 percent of peak prices.

DeepSeek's technical details are the company's own and have not been independently tested. For readers, that means the promises of low prices and high performance should be treated with caution until others have tried the service.