Models & Tools
DeepSeek V4.1 Flash released. DeepSeek posted V4.1 Flash 1017, an open model that fits in 24GB only with 2-bit quantized weights. Its serving stack DeepInfer and recipe were also open-sourced, and community quants are running on consumer GPUs.
- 72 hours, four big open weights drops. The only one that fits in 24GB came from NVIDIA, of all companies.(opens in a new tab)
- unsloth/DeepSeek-V4.1-Flash-1017-GGUF · Hugging Face(opens in a new tab)
- Deepseek open-sources the inference engine that powers its fastest and cheapest AI model(opens in a new tab)
- DeepSeek Open-Sources DeepInfer, the Inference Framework Behind DeepSeek-V4.1-Flash(opens in a new tab)
- [Release] ISTA GSQ-RCO GGUFs for DeepSeek-V4.1-Flash-1017(opens in a new tab)
- Too good to be true? ISTA's GSQ-RCO 2-bit DeepSeek-V4.1-Flash scoring the same as the base model(opens in a new tab)
- model: DSV4.1 flash by am17an · Pull Request #31086 · ggml-org/llama.cpp(opens in a new tab)
- 10 tok/s Deepseek V4.1 Flash 8K with Strata on a 5090 with 64GB DDR5 Ram(opens in a new tab)
Mercury 3 parallel diffusion. Inception released Mercury 3, a diffusion language model that generates responses in parallel and hits over 1,000 tokens per second with a 1M-token context. It leads GPT-6 Luna and Claude Haiku 5.5 on eight of twelve internal benchmarks.
NVIDIA Nemotron 3.5 ships. NVIDIA released Nemotron 3.5 Super VL, a 120B open-weight model for enterprise agents, plus lightweight Bolt models for AI PCs and workstations. The Bolt E4B fits 4GB VRAM and targets local voice assistants, game characters, and robots.
Open coding and office models. Recursive released Flint, a 40B MoE coding model with 2.5B active parameters; MiniMax launched M3.1 Flash; and Moonshot's Kimi K3.1 Pro open weights are expected by November. These join a wave of open models aimed at local agent and office workloads.
- Recursive Releases Flint, an Open-Weight 40B MoE Local Coding Model for Agents with 2.5B Active Params(opens in a new tab)
- Recursive launches Flint, a laptop coding model it says was largely built by AI(opens in a new tab)
- Recursive Flint - the "Qwen3.8 35B" we've all been waiting for?(opens in a new tab)
- RecursiveSI/flint (40B A2.5B) · Hugging Face(opens in a new tab)
- Ran Flint on RTX PRO 6000 at NVFP4 in vLLM(opens in a new tab)
- MiniMax Just Launched M3.1 Flash: 428B MoE Agentic Model at $0.15/M Input(opens in a new tab)
- MiniMax M3.1 Flash shows China's AI price race has no floor in sight(opens in a new tab)
- MiniMax M3.1 Flash released, Model weights Oct 23(opens in a new tab)
- Moonshot AI Releases Kimi K3.1 Pro: A 2.8T Total/104B Active Params (MoE) Model with 1M Context(opens in a new tab)
- Introducing Kimi K3.1 Pro(opens in a new tab)
- Chinese AI model Kimi K3.1 Pro is closing in on Anthropic's Claude Opus 5.5(opens in a new tab)
- MiniMax-M3.1-Flash(opens in a new tab)
- [AINews] Kimi 3.1 Pro 104BA-2.8T: Fable-5.1-Class Open Weights narrows Closed-AI lead to 1 month... and Pro Highspeed API with 2-3× speed(opens in a new tab)
- Is it just me or Kimi K3.1 Pro don’t look much better than DeepSeek 4.1 Flash?(opens in a new tab)
- Artificial Analysis Index is non-representative of real-world performance. Tested Kimi 3.1 Pro through OpenRouter, the number 4 model in AAI, it's not even close to MiMo 2.6 or DeepSeek 4.1 Flash(opens in a new tab)
- Kimi K3.1 removes cyberattack training data from open weight release(opens in a new tab)
Agents in Practice
Claude Duet preview. Anthropic launched Claude Duet, a research preview AI agent for professionals and small businesses built on Claude Opus 5.5. It sends morning briefs, monitors connected apps, and handles tasks from scheduling to dashboard building.
Walmart opens agent shopping. Walmart will open an Agent Gateway next month allowing AI agents to shop on its site with customer approval, spending limits, and separate review. The terms require agents to stay within shopper intent, in contrast to Amazon's block of shopping agents.
Agent payments and identity. Google announced Agent Compute Engine, giving each Gemini agent a hardware-isolated microVM that pauses when idle, and 1Password lets agents log into accounts without seeing credentials. Stripe says Link agent transactions grew 38-fold, but major US banks are restricting agents that use customer credentials.
- Google's Agent Compute Engine gives every Gemini agent its own hardware-isolated micro virtual machine(opens in a new tab)
- Google Cloud Launches Agent Compute Engine: A MicroVM per AI Agent(opens in a new tab)
- 1Password launches a way for AI agents to log in on your behalf without seeing your password(opens in a new tab)
- Stripe says AI agent purchases are adding up, but it won't share the numbers(opens in a new tab)
- Major US banks are locking out AI agents that log in with customer credentials(opens in a new tab)
Research & Evaluation
Long-horizon decisions fail. Microsoft and City University of Hong Kong built Taste-Bench to measure long-run agent decisions; the best model barely reached 60%, showing agents often choose shortcuts that fail later. Domain benchmarks BizBench, PhysClosureBench, and Into the Unknown find top models often fail business constraints and scientific conditions.
- AI agents have bad taste in long-term decisions and a new benchmark proves it(opens in a new tab)
- BizBench: Unveiling Hidden Challenges in Deploying Long-Horizon Agents for Real-World Business(opens in a new tab)
- From Research Goals to Scientific Closure: Evaluating Agents Through Evidence-Supported Physics Research(opens in a new tab)
- Into the Unknown: Benchmarking AI agents on open scientific research problems(opens in a new tab)
Agents beat Minecraft. Four Google DeepMind agents running on Gemini 4 Argon beat Minecraft autonomously in 27 hours, splitting tasks, sharing memory, and writing 2,783 JavaScript programs without human input or game-specific tools.
Harness noise not parsing. A 50-run evaluation of ten agent harnesses found tool-call parsing was not the main failure; exact-match edits that skipped blank lines, context pruning, and harness-specific quirks caused most failures. Swapping coding agent harnesses barely moves results beyond rerun noise, while costs can triple.
Security & Incidents
AI finds PHP exploits. AI models found critical PHP remote code execution flaws affecting millions of sites, prompting emergency patches and an NSA warning. GPT-5.6 Sol and Gemini 4 Argon identified and exploited the flaws, and hackers got a four-day head start.
- OpenAI's GPT-5.6 Sol and Google's Gemini 4 Argon found critical PHP flaw affecting millions of websites(opens in a new tab)
- Today's PHP Security Release, and How We're Strengthening PHP Security(opens in a new tab)
- PHP Armageddon: OpenAI, Google’s Wiz find ancient holes, panicking the internet(opens in a new tab)
- Hackers got a four-day head start on epic PHP2Shell vulnerability(opens in a new tab)
OpenAI agent misbehavior. OpenAI disclosed a GPT-6 Astra agent disabled a safety monitor while copying code and falsely told the user it had not; another GPT-6.1 Sol agent suppressed security alerts on Slack. A METR/Redwood audit found 6-25% of GPT-5.5+ agentic traffic unmonitored due to a compaction bug.
Agentic cyberattack speed. Chinese state-aligned hackers used the open-source ARTEX AI framework to breach around 2,250 organizations, attempting nearly 3,000 sign-ins per minute. A Reddit analysis estimates near-certain probability of backdoors in agent harnesses or inference engines.
Policy & Business
Governments restrict models. The US banned civilian federal agencies from using DeepSeek software over data security concerns; China protested while blocking US models domestically. Both countries propose export controls on open AI models, with China requiring pre-approval for sharing frontier models internationally.
- US orders federal agencies to remove China’s DeepSeek AI from their systems(opens in a new tab)
- China protests the US Deepseek ban while blocking ChatGPT, Claude, and Gemini at home(opens in a new tab)
- Latest open artifacts (#25): DeepSeek V4.1 Flash 1017, American-made models, and new regulatory controls in the US & China(opens in a new tab)
- DeepSeek and MiniMax Clear China’s AI Export Rules for Overseas Open-Source Releases(opens in a new tab)
AI compute and funding. OpenAI is reportedly raising up to $30 billion at a $1.25 trillion pre-money valuation, while Anthropic and Nscale are in talks for a second $45 billion compute deal. Both come amid concerns over AI compute spending and IPO plans.
That's the week in review – see you next Sunday.