DeepSeek launches V4. 1 Flash multimodal AI model

DeepSeek V4. 1 Flash Mixture-of-Experts multimodal AI AI API pricing open-weight models

DeepSeek has released V4. 1 Flash, a new Mixture-of-Experts (MoE) model designed for high performance, faster inference, and lower deployment costs. The model features a new Causal Encoder-Decoder architecture and native multimodal support, allowing it to process both text and images.

Technically, the model utilizes 552 billion total parameters but only activates 8 billion during input processing and 16 billion during output generation. This efficiency is bolstered by a reduction in KV cache requirements, which reportedly lowers high-bandwidth memory (HBM) demand to one-fourth and SSD storage needs to one-eighth compared to previous versions. The model supports a 1 million-token context window and up to 384,000 output tokens.

In terms of commercial strategy, DeepSeek has lowered Flash-series API pricing and announced that all existing V4-Pro requests will be automatically routed to V4. 1 Flash at the lower rate until a successor is released. Benchmarks suggest the model competes with or exceeds flagship models from Anthropic and OpenAI in specific agentic and coding tasks. Additionally, reports indicate DeepSeek has engaged CITIC Securities to prepare for an IPO on the STAR Market.

Entities

Anthropic · CITIC Securities · DeepSeek · Hugging Face · OpenAI

Claims

What the coverage asserts, and how many sources carry each claim.

[● 2 SOURCES] V4. 1 Flash reduces high-bandwidth memory (HBM) demand to one-fourth and SSD demand to one-eighth compared to the previous generation. unwire.pro · time.news
[DISPUTED] The model has 552 billion total parameters, activating 8 billion for input and 16 billion for output. unwire.pro · thebridge.jp · time.news · gigazine.net
[○ 1 SOURCE] The Artificial Analysis Intelligence Index v4.3 gave the model a score of 40, surpassing GPT-5.6 Luna. gigazine.net
[○ 1 SOURCE] DeepSeek has reportedly engaged CITIC Securities to prepare for an initial public offering (IPO) on the STAR Market. technews.tw
[● 5 SOURCES] The V4. 1 Flash model features a new architecture with native multimodal support for both text and images. pakistanpost.pk · unwire.pro · thebridge.jp · time.news · gigazine.net
[● 4 SOURCES] DeepSeek claims V4. 1 Flash outperforms the V4-Pro model in performance, cost, speed, and total completion time. pakistanpost.pk · unwire.pro · www.ithome.com · www.tmtpost.com
[● 2 SOURCES] The model supports a 1 million-token context window and a maximum output of 384,000 tokens. thebridge.jp · time.news
[● 4 SOURCES] DeepSeek will automatically route V4-Pro API requests to V4. 1 Flash at the lower Flash pricing until a V4. 1 Pro model is released. pakistanpost.pk · unwire.pro · www.ithome.com · thebridge.jp