DeepSeek: latest AI news
All Oossa news about DeepSeek, newest first.
Known for: DeepSeek V4
Follow via RSSToday
Models
AI21 says harness changes cut GitLab review costs 97.7%
AI21 tested GitLab Duo Code Review on 50 public pull requests. DeepSeek V4 Flash and compacted prompts cut measured cost while keeping benchmark F1 near baseline.
Business
DeepSeek secures $15 billion in new funding round
China's AI firm DeepSeek raised about $15 billion, fueling plans for overseas growth and a possible IPO.
Yesterday
Models
DeepSeek adds Huawei Ascend support to six AI repos
DeepSeek released Ascend hardware support for TileLang, DeepGEMM, DeepEP, TileKernels, FlashMLA and DeepSelect on Sept. 30, 2026.
Safety
Hackers use older AI models to bypass safeguards in attacks
CrowdStrike reports four groups exploiting open AI tools like Anthropic Claude and Deepseek to run autonomous hacks between May and August 2026.
Policy & law
US bans Deepseek AI products; China says it’s discrimination
The US Department of Homeland Security ordered federal agencies to drop Deepseek tools, while China blocks ChatGPT, Claude and Gemini at home.
Tuesday, October 20
Models
DeepSeek releases DeepInfer inference framework as open source
DeepSeek has made its DeepInfer framework, used for the DeepSeek‑V4.1‑Flash model, available under an Apache 2.0 license.
Safety
US bans DeepSeek AI tools in federal civilian agencies
The Department of Homeland Security ordered all civilian agencies to remove DeepSeek products on Oct. 19, 2026, citing security threats.
Saturday, October 17
Models
OpenRouter adds DeepSeek V4.1 Flash 1017 model
OpenRouter now offers DeepSeek V4.1 Flash 1017, a sparse mixture‑of‑experts model with a 1‑million token context window.
Models
DeepSeek releases V4.1-Flash-1017 model on Hugging Face
The new DeepSeek-V4.1-Flash-1017 model improves reasoning, coding and visual tasks while cutting runtime and output size.
Tuesday, October 6
Models
DeepSeek releases open-weight V4.1-Flash for coding agents
The new model uses fewer parameters for input processing, cutting cost and memory for agentic workflows.
Models
Fireworks adds visual support to DeepSeek-V4-Flash model
The experimental DeepSeek-V4-Flash-Vision-Exp model now handles images, with on‑demand GPU deployment and a deprecation notice for serverless use.
Business
Deepseek raises $12‑$15 billion, eyes 2027 IPO
Chinese AI startup Deepseek is close to a $12‑$15 billion funding round backed by CAT and Tencent, and plans an IPO in early 2027.
Sunday, October 4
Friday, October 2
Thursday, October 1
Wednesday, September 30
Models
Deepseek releases open‑source tools for Huawei Ascend AI chips
Deepseek announced TileLang and related libraries as open‑source software for Huawei's Ascend 950 chips, aiming to simplify AI development in China.
Models
DeepSeek releases LibreOffice‑kit 0.1.0 for Node.js document conversion
The new @deepseek-ai/libreoffice-kit 0.1.0 adds PDF conversion, spreadsheet recalculation and PNG rendering, with smaller native and WASM packages.
Models
DeepSeek releases DeepEP-Ascend library for Huawei NPU communication
DeepSeek open‑sources a high‑performance communication library that speeds up MoE training on Ascend NPUs, with early benchmarks showing up to 95% bandwidth utilization.
Models
DeepSeek releases DeepGEMM-Ascend for Huawei Ascend NPU
The new repo ports DeepGEMM to Ascend, supporting BF16, FP8, FP4 GEMM and MegaMoE with near‑peak hardware performance.
