DeepSeek

A developer extended Sebastian Raschka's GPT-2-style code from the book "Build a Large Language Model (from Scratch)" to add mixture-of-experts support and trained a 446M-parameter model with 220M act…

DeepSeek launched DeepSeek-V4.1-Flash on Thursday, an open-weights model with 552 billion total parameters that activates 8 billion parameters per token for input and 16 billion for generation, releas…

The NSA, CISA, and FBI have accused six Chinese firms — DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI — of running "industrial-scale" distillation attacks on US models including Claude, G…

DeepSeek released DeepSeek V4.1 Flash on 2026-09-10, a text-and-image model with a 1000k context window and 384k maximum output, according to the provider announcement. No retirement date has been ann…

Anthropic PBC accused Moonshot AI of covertly routing thousands of user requests to Anthropic's Claude models and passing off the responses as its own, according to a report Anthropic published Thursd…

Nari Labs launched a Free Public Beta of realtime text-to-speech endpoints built on its own inference engine for the Qwen3-TTS model, claiming 50 ms time to first audio in Fast mode versus 300 ms for …

DeepSeek introduced its DeepSeek-V4.1-Flash model into the Agent Arena for testing on complex real-world tasks, evaluated via a causal tracing methodology with scores expected soon. Prediction-market …

ByteDance's Seed team, working with University of Hong Kong researchers, adapted DeepSeek's Group Relative Policy Optimization (GRPO) technique into a visual-generation framework called DanceGRPO, rep…

DeepSeek launched V4.1-Flash on September 10, a 552-billion-parameter Mixture-of-Experts model with a 1-million-token context window that activates roughly 8 billion parameters for input and 16 billio…

DeepSeek released DeepSeek V4.1 Flash on September 10, 2026, a sparse mixture-of-experts model the company reports exceeds its V4 Pro flagship on performance, speed, and task completion time. The mode…

A developer released codsh 0.22.1, a DeepSeek-native terminal coding agent that turns a one-sentence request via its /ship command into verified code. The update fixes grill cards so long questions wr…

DeepSeek released its V4.1 Flash model on Thursday, claiming it outperforms its previous flagship V4 Pro on coding, cybersecurity and autonomous agent benchmarks while cutting inference costs. The mod…

DeepSeek released V4.1 Flash on September 10, 2026, cutting cache-hit costs to $0.003 per token during off-peak hours from the $0.022 charged for the outgoing V4-Pro, a 77-80% price reduction, and rai…

DeepSeek released DeepSeek V4.1 Flash, a 552B-parameter MoE model built on a new Causal-Encoder-Decoder architecture with 8B activated parameters on the input side and 16B on the output side, availabl…

DeepSeek released version 0.1.5 of DeepSeek Harness on Sept. 10, adding dedicated optimization for V4.1 Flash plus file-upload and sidebar-preview features, according to IT Home. Version 0.1.5 support…

A developer outlined a new AI engineering approach called "Agentic Harnessing," which layers tools, memory, and an agentic loop on top of base AI models to turn them into context-aware agents. The wri…

DeepSeek released the V4.1 Flash model on September 10, 2026 (Beijing Time) and postponed the discontinuation of its V4 Pro service to 12:00 Beijing Time on September 14, 2026, according to an email s…

DeepSeek released V4.1-Flash, an efficiency-focused model with 552 billion total parameters (8 billion active for input, 16 billion for output) that matches or beats GPT-5.6 Sol and Claude Opus 5 on s…

ZenMux launched a unified AI model gateway that offers built-in insurance compensation when outputs are hallucinated, excessively slow, or low-throughput, according to the company's product page. The …

DeepSeek released DeepSeek V4.1 Flash, a 552B-parameter mixture-of-experts model built on a new causal encoder-decoder architecture that natively supports multimodal visual understanding. The model is…