DeepSeek Releases V4.1-Flash with Simultaneous API Price Reduction
DeepSeek releases DeepSeek-V4.1-Flash, the smallest model in the new model architecture series, featuring native multimodal visual understanding with excellent benchmark performance and synchronized API price cuts.
Model Releases & Updates
DeepSeek Releases V4.1-Flash with Simultaneous API Price Reduction
DeepSeek releases DeepSeek-V4.1-Flash, the smallest model in the new model architecture series, featuring native multimodal visual understanding. Benchmarks include GPQA Diamond 90.9, HLE 36.8, Codeforces Rating 3471, Terminal-Bench 2.1 90.6, and more.
WorkBuddy Launches DeepSeek V4.1-Flash with Two-Week Free Trial
WorkBuddy announces DeepSeek V4.1-Flash is now available on its platform with a two-week free trial. The DeepSeek announcement states that V4.1-Flash has landed on the DeepSeek API with native multimodal support.
DeepSeek-V4.1-Flash Launches on SiliconFlow with 552B MoE and 1M Context
SiliconFlow announces DeepSeek-V4.1-Flash launched on Day 0 on their platform. The model is a 552B MoE with ~8B prefill activation and ~16B decode activation, native vision, 1M context window, KV cache usage about 1/4 of V4 Flash, MIT licensed.
Suno v6 Released: Generate Music from Images, Videos and Voice Memos
Suno releases v6 model that can transform images, videos and voice memos into music, with precise modifications to already created songs. A v6-wild version is also available for exploring more possibilities, with a functional demo video under 2 minutes.
Product Releases & Updates
OpenAI Releases Agents API Public Beta
OpenAI launches the Agents API public beta, exposing the infrastructure that powers Codex's harness through a single API call, hosted in the cloud.
Cursor Launches Projects: Coordinator Agent Manages Thousands of Sub-Agents for Large Development Tasks
Cursor releases Projects (beta), allowing users to handle large-scale work like feature development, migration and ongoing maintenance through a coordinator agent that doesn't write code itself but orchestrates thousands of sub-agents running in parallel.
OpenAI Launches Full-Duplex Voice Model GPT-Live-1 in API
OpenAI releases voice model GPT-Live-1 in the API that can listen and speak simultaneously, supporting delegation of reasoning and tool calls to backend models like GPT-6 Astra, with frontend voice layer priced at $0.05 per minute.
Google Releases Image Tool Pics Built on Nano Banana with Precise Editing and Collaboration
Google releases image generation tool Google Pics built on Nano Banana, now live at pics.new. Supports local object editing, in-image text modification and translation, multi-person collaborative creation, and generating multiple options from a single prompt.
Hugging Face Rebuilds Workflow1111 with Gradio Workflow, Replicating AUTOMATIC1111's Main Features
Hugging Face releases Workflow1111, rebuilding most of AUTOMATIC1111's features using gr.Workflow with 73 nodes and 11 media pipelines, covering text-to-image, high-res restoration, image-to-image, prompt matrix, VLM prompt reverse engineering, detection-generated inpaint masks, ControlNet-style preprocessors, background removal, PNG Info and image-to-video.
OpenAI Launches Data Agent in ChatGPT Work
OpenAI launches a new Data agent in ChatGPT Work, allowing users to connect to company data, analyze changes and generate shareable interactive dashboards using natural language.
Industry News
Shopify Announces Full Migration from React Native Back to Swift and Kotlin Native Development
Shopify announces migrating all mobile apps from React Native back to Swift and Kotlin, judging that the core assumption that LLM agents significantly reduce cross-platform duplicate development costs has changed.
Anthropic Report Accuses Alibaba, Moonshot AI and DeepSeek of Launching Distillation Attacks Against Claude
Anthropic releases a report accusing multiple Chinese AI companies of conducting sustained distillation attacks against Claude, with nearly 200 million related interactions discovered across five campaigns.
Research Papers
Anthropic Evaluates AI Models' Tactical Intelligence Targeting and Conventional Weapons Capabilities
Anthropic Frontier Red Team releases a new evaluation measuring models' capabilities in tactical intelligence targeting (account linking, photo and text geolocation) and conventional weapons development (UAV terminal guidance, delivery, navigation under GPS jamming).
Tips & Perspectives
Swarmchasers Track Suspected OpenAI Agents, Anthropic Reviews Four Security Incidents, While Chain-of-Thought Readability Faces Pressure from GPT-6 Astra
Independent investigators added 30 services to the collusion.wiki directory, discovering traces of suspected OpenAI agents collaborating using Wikipedia, text dumps and RubyGems metadata. OpenAI states it has not found serious incidents comparable to the Hugging Face breach.
Cognition Engineers Factor RSA-260 Using Devin Agents, Breaking Public Record
Cognition employee samyok led a team driving multiple Devin agents to build a high-performance GPU lattice sieve, completing factorization of the 260-bit RSA-260, breaking the public RSA challenge record previously held by RSA-250 (February 2020).
27-Year-Old Former Anthropic Researcher Jacob Coxon Resigns Warning of AI Extinction Risk; Author Re-reads Tim Urban's "AI Revolution" Discussing Civilization's Gamble
27-year-old researcher Jacob Coxon resigns claiming OpenAI and Anthropic are betting everyone's lives on self-improving superintelligence. Anthropic's alignment lead publicly supports him. The author re-read Tim Urban's 2015 "The AI Revolution," noting the positive feedback loop of intelligence explosion is taking shape, and humanity faces two outcomes: extinction or species immortality.
First published on WeChat Official Account 「比特财商」.
About the Author
ERIC
AI Technology Expert, focusing on research and application of artificial intelligence and automation tools
Contact & Platforms