https://tpsreport.news/news/anthropic-self-replicating-ai-botnets-experts-respond TPS — Tokens Per Second en 2026-10-31T13:05:51.266Z Anthropic Warns Small AI Models Are Months From Self-Replicating Botnets; Experts Call Threat Familiar Anthropic, AI safety, botnets, self-replication, cybersecurity, Mirai, Frontier Red Team, open-weight models https://tpsreport.news/news/claude-opus-5-5-game-recreation-reverse-engineering-refusals TPS — Tokens Per Second en 2026-10-31T12:35:46.240Z Users report Claude refusing game recreation and reverse-engineering tasks; Anthropic silent Anthropic, Claude Code, Claude Opus 5.5, reverse engineering, Cyber Verification Program, Project Glasswing, safety policy, game development https://tpsreport.news/news/factory-droid-1-0-model-independent-coding-agent-harness TPS — Tokens Per Second en 2026-10-31T12:20:46.140Z Factory releases Droid 1.0 agent harness, 65.2% on Terminal-Bench 4 with Claude Opus 5.5 Factory, Droid, coding agent, Terminal-Bench 4, Claude Opus 5.5, agent harness https://tpsreport.news/news/deepmind-forty-global-forest-type-map-alphaearth TPS — Tokens Per Second en 2026-10-31T10:35:45.282Z DeepMind and WRI release ForTy, a 10-meter global forest type map with 90.2% overall accuracy Google DeepMind, ForTy, AlphaEarth Foundations, World Resources Institute, satellite imagery, forest mapping, EUDR, Google Earth Engine https://tpsreport.news/news/china-ai-labs-recursive-self-improvement-without-weights TPS — Tokens Per Second en 2026-10-31T07:35:50.184Z Chinese AI labs aim self-improvement at training infrastructure, not model weights, Leiphone survey finds recursive-self-improvement, Zhipu, DeepSeek, Moonshot AI, Qwen, ModelBest, ByteDance Seed, Xiaomi https://tpsreport.news/news/ace-robotics-ntu-s-lab-mundo-world-model-photos-robot-training TPS — Tokens Per Second en 2026-10-31T07:20:52.034Z ACE Robotics and NTU S-Lab's Mundo builds simulated robot training scenes from 1 to 12 photos Mundo, world model, ACE Robotics, NTU S-Lab, robotics, MuJoCo, real-to-sim, video diffusion https://tpsreport.news/news/google-pagebreak-ai-security-agent-500-xss-bugs TPS — Tokens Per Second en 2026-10-31T06:35:45.263Z Google's PageBreak AI Agent Has Found 500+ XSS Bugs, Reporting Only Exploits Validated in Running Apps Google, PageBreak, Gemini, AI security, XSS, vulnerability research, CodeMender, agents https://tpsreport.news/news/qoder-desktop-interconnect-remote-coding-tasks-no-ssh TPS — Tokens Per Second en 2026-10-31T03:35:40.792Z Alibaba's Qoder adds desktop interconnect to run coding tasks on another machine without SSH Qoder, Alibaba, AI coding, remote development, desktop client, developer tools https://tpsreport.news/news/xiaomi-ocr-0-open-source-0-8b-omnidocbench TPS — Tokens Per Second en 2026-10-31T03:06:32.358Z Xiaomi open-sources Xiaomi-OCR-0, a 0.8B document model scoring 96.83 on OmniDocBench v1.6 Xiaomi, Xiaomi-OCR-0, OCR, document parsing, open source, OmniDocBench, small models, Hugging Face https://tpsreport.news/news/baidu-search-recursive-self-evolution-api-4-9-yuan-per-1000-calls TPS — Tokens Per Second en 2026-10-31T03:06:03.787Z Baidu Cuts Search AI API to 4.9 Yuan per 1,000 Calls, Claims Search Now Self-Evolves Baidu, search API, pricing, MCP, DeepSeek, AI search, Baidu AI Cloud https://tpsreport.news/news/deepseek-cyber-cybersecurity-model-limited-testing TPS — Tokens Per Second en 2026-10-31T03:05:41.892Z DeepSeek website code references Cyber, a cybersecurity model with identity-gated API access, in limited testing DeepSeek, Cyber, cybersecurity, cyber-personal-v3, cyber-enterprise-v4, API access, limited testing https://tpsreport.news/news/apple-intelligence-china-ios-27-3-alibaba-qwen TPS — Tokens Per Second en 2026-10-30T23:35:44.211Z Apple Intelligence reaches China with iOS 27.3 in December, with Alibaba's Qwen inside Siri Apple, Apple Intelligence, Alibaba, Qwen, China, iOS 27.3, Siri https://tpsreport.news/news/openai-chatgpt-teens-weapons-information-limits TPS — Tokens Per Second en 2026-10-30T21:05:36.308Z OpenAI tightens ChatGPT limits on weapons-related information for teens, even for harmless-looking questions OpenAI, ChatGPT, teen safety, safeguards, content policy, Reuters https://tpsreport.news/news/ai-security-agents-compared-claude-codex-mantis-cloudflare TPS — Tokens Per Second en 2026-10-30T19:35:58.670Z Anthropic, OpenAI, Cloudflare and Google each release AI agents that hunt security bugs in code security, AI agents, code security, Anthropic, OpenAI, Cloudflare, Google, MarkTechPost https://tpsreport.news/news/anthropic-claude-max-ultra-plan-rename-5-hour-limit-removed TPS — Tokens Per Second en 2026-10-30T19:35:40.300Z Anthropic renames Max 5x to Max and Max 20x to Ultra, drops 5-hour limit for verified Ultra accounts Anthropic, Claude, Claude Max, Claude Ultra, pricing, subscription plans, usage limits https://tpsreport.news/news/liquid-ai-lfm3-1-7b-gguf-qad-q4-0 TPS — Tokens Per Second en 2026-10-30T19:06:06.539Z Liquid AI publishes LFM3-1.7B GGUF with QAD Q4_0 build at 1.05 GB for local CPU and GPU inference Liquid AI, LFM3, GGUF, local inference, quantization, llama.cpp, Ollama, LM Studio https://tpsreport.news/news/anthropic-claude-five-exploit-chains-browsers-operating-systems TPS — Tokens Per Second en 2026-10-30T18:20:44.778Z Anthropic Says Unreleased Claude Model Built Five Working Exploit Chains Across Chrome, Safari, Android, Windows and Lin Anthropic, Claude, cybersecurity, exploit chains, vulnerability research, AI safety https://tpsreport.news/news/openai-model-spec-five-core-values-shutdown TPS — Tokens Per Second en 2026-10-30T17:50:44.228Z OpenAI rebuilds its Model Spec around five core values, bars GPT from resisting shutdown OpenAI, Model Spec, AI safety, AI agents, GPT, alignment https://tpsreport.news/news/project-suncatcher-orbiting-tpu-gemma-4-haiku TPS — Tokens Per Second en 2026-10-30T17:35:50.860Z Google's Project Suncatcher satellite returns first AI response from orbit: a Gemma 4 haiku run on a TPU Google, Project Suncatcher, Gemma 4, TPU, space computing, Google Research https://tpsreport.news/news/nvidia-nemoclaw-open-agent-blueprints-nemotron-openshell TPS — Tokens Per Second en 2026-10-30T16:20:56.505Z NVIDIA NemoClaw blueprints bundle Nemotron, NeMo tools and OpenShell for three agent harnesses NVIDIA, NemoClaw, Nemotron 3 Ultra, OpenShell, AI agents, LangChain, OpenClaw, Hermes Agent https://tpsreport.news/news/agibot-go-3-preview-wita-omni-embodied-benchmarks TPS — Tokens Per Second en 2026-10-30T15:07:26.032Z AgiBot launches GO-3 Preview and WITA-OMNI, topping three and four embodied AI leaderboards AgiBot, embodied AI, WITA-OMNI, GO-3 Preview, robotics, Lingqu OS, physical AI https://tpsreport.news/news/forgekernel-beats-deepseek-flashmla-prefill-b200 TPS — Tokens Per Second en 2026-10-30T15:06:11.517Z ForgeKernel agent beats DeepSeek's FlashMLA Prefill kernel by 9.8% on Nvidia B200 with zero human-written code ForgeKernel, DeepSeek, FlashMLA, Mianbi Intelligence, OpenBMB, Nvidia B200, Ascend, CANNBench https://tpsreport.news/news/microsoft-copilot-synthetic-enterprise-data-evaluation TPS — Tokens Per Second en 2026-10-30T14:36:03.976Z Microsoft builds synthetic enterprise tenants to evaluate Copilot without using customer data Microsoft, Copilot, evaluation, synthetic data, enterprise AI, research https://tpsreport.news/news/altman-openai-math-hundreds-of-millions-spending TPS — Tokens Per Second en 2026-10-30T13:20:56.742Z Altman says OpenAI has spent hundreds of millions on math and will keep releasing results OpenAI, Sam Altman, mathematics, AI for science, Navier-Stokes, Cheeky Pint, Yang-Mills https://tpsreport.news/news/microsoft-mage-vl-cowork-4b-computer-use-osworld TPS — Tokens Per Second en 2026-10-30T12:22:12.051Z Microsoft releases Mage-VL-Cowork, a 4B computer-use agent scoring 66.0% on OSWorld-Verified Microsoft, Mage-VL-Cowork, computer-use, OSWorld-Verified, open-weights, vision-language, 4B, agents https://tpsreport.news/news/tbc-mischief-wan-2-2-image-to-video-fal-ai TPS — Tokens Per Second en 2026-10-30T12:21:13.061Z TBC launches Mischief, a Wan 2.2-based image-to-video model, on fal.ai October 30 TBC, Mischief, Wan 2.2, image-to-video, fal.ai, video generation https://tpsreport.news/news/samaya-ai-sterling-finance-research-model-frontierfinance TPS — Tokens Per Second en 2026-10-30T09:35:45.514Z Samaya AI's Sterling scores 58.9% on FrontierFinance at $0.61 per question, ahead of Claude Fable 5.1 Samaya AI, Sterling, FrontierFinance, Nemotron 3.5 Lightning, Claude Fable 5.1, finance, agents https://tpsreport.news/news/gemini-notebook-free-textbooks-openstax-s-chand TPS — Tokens Per Second en 2026-10-30T08:50:42.775Z Google adds hundreds of free textbooks as sources in Gemini Notebook, available from Japan Gemini Notebook, Google, education, textbooks, OpenStax, S Chand Publishing, Expert Intelligence https://tpsreport.news/news/open-h3-regenerate-2k-nvidia-mit-4k-video TPS — Tokens Per Second en 2026-10-30T08:21:01.067Z NVIDIA and MIT team releases Open-H3-Regenerate-2K, regenerating MiniMax H3 video drafts at up to native 4K Open-H3-Regenerate-2K, MiniMax H3, NVIDIA, MIT, video generation, open weights, 4K, Hugging Face https://tpsreport.news/news/pika-h3-flex-open-weight-video-six-steps TPS — Tokens Per Second en 2026-10-30T07:51:07.433Z Pika releases H3 Flex, an open-weight video model that generates synchronized audio in six sampling steps Pika, Pika H3 Flex, video generation, open weights, Hugging Face, audio generation, reference-guided video https://tpsreport.news/news/edge0-audio8-tts-realistic-0-6b-voice-cloning TPS — Tokens Per Second en 2026-10-30T05:35:43.059Z Edge0 releases Audio8-TTS-Realistic, a 0.6B open-weight voice cloning TTS model under Apache 2.0 Edge0, Audio8-TTS-Realistic, TTS, voice cloning, open-weight, Apache 2.0, GRPO https://tpsreport.news/news/sarashina3-2-mini-thinking-sb-intuitions-api TPS — Tokens Per Second en 2026-10-30T02:50:52.979Z SB Intuitions launches Sarashina3.2 mini thinking, the Sarashina series' first reasoning model, via API Sarashina, SB Intuitions, SoftBank, reasoning model, thinking model, Japanese LLM, API, tool calling https://tpsreport.news/news/tencent-hunyuan-hy-nomas-technical-report-hy-nomas-bench TPS — Tokens Per Second en 2026-10-30T02:21:56.259Z Tencent Hunyuan publishes Hy-Nomas report, releases 280-case world model benchmark under Apache 2.0 Tencent Hunyuan, Hy-Nomas, Hy-Nomas-Bench, world models, open source, benchmark, video generation, Apache 2.0 https://tpsreport.news/news/agibot-go-3-preview-wram-embodied-model-open-source TPS — Tokens Per Second en 2026-10-30T02:21:31.367Z AgiBot launches GO-3 Preview embodied model, plans to open-source it by end of 2026 AgiBot, GO-3, embodied AI, robotics, WRAM, open source, RoboDojo, RoboTwin https://tpsreport.news/news/doubao-work-kimi-k3-1-pro-glm-5-3-deepseek-v4-1-flash-enterprise TPS — Tokens Per Second en 2026-10-30T02:20:54.323Z Doubao Work adds Kimi K3.1 Pro, GLM-5.3 and DeepSeek-V4.1-Flash for team and enterprise plans Doubao Work, ByteDance, Kimi K3.1 Pro, GLM-5.3, DeepSeek-V4.1-Flash, enterprise AI, AI agents, Moonshot AI https://tpsreport.news/news/glm-5-3-firefox-patch-gap-epoch-ai TPS — Tokens Per Second en 2026-10-29T23:20:46.299Z Epoch AI: Open-Weight GLM 5.3 Exploits Firefox Patch Gap at About $1,000 per Version GLM 5.3, Z.ai, Epoch AI, Firefox, cybersecurity, open-weight, patch gap https://tpsreport.news/news/tinyllama-medusa-2-knowledge-distillation-booking-com TPS — Tokens Per Second en 2026-10-29T22:35:57.305Z AWS and Booking.com shrink Llama-3.1 70B to 1.1B TinyLlama, 10-20x faster at matched quality knowledge-distillation, Medusa, TinyLlama, Llama-3.1, Booking.com, Amazon, AWS, inference-optimization https://tpsreport.news/news/openai-doubles-pro-plan-usage-limits-work-codex TPS — Tokens Per Second en 2026-10-29T20:20:42.225Z OpenAI doubles Work and Codex quotas on Pro 100 and Pro 500 plans OpenAI, usage limits, Pro plans, Codex, Anthropic, ChatGPT https://tpsreport.news/news/gemini-app-edit-documents-pdfs-in-chat TPS — Tokens Per Second en 2026-10-29T19:20:59.653Z Gemini app adds in-chat editing, full-screen preview and text highlighting for generated files Gemini, Google, Gemini app, document editing, PDF, product update https://tpsreport.news/news/github-copilot-repository-plugin-defaults-enterprise-controls-ga TPS — Tokens Per Second en 2026-10-29T19:20:44.905Z GitHub makes repository plugin defaults and enterprise plugin controls generally available in Copilot GitHub Copilot, plugins, VS Code, Copilot CLI, enterprise, GitHub https://tpsreport.news/news/baseten-joey-1-2b-open-weights-decision-model TPS — Tokens Per Second en 2026-10-29T18:06:18.662Z Baseten releases joey-1, a 2B-parameter open-weights decision model, at $0.01 per million tokens Baseten, joey-1, Base Labs, open weights, decision model, Hugging Face https://tpsreport.news/news/elevenlabs-sfx-v3-sound-effects-model TPS — Tokens Per Second en 2026-10-29T18:05:56.308Z ElevenLabs releases SFX v3, a sound effects model built for multi-sound prompts ElevenLabs, SFX v3, sound effects, audio generation, ElevenCreative, ElevenAPI https://tpsreport.news/news/openai-expands-frontier-model-misalignment-monitoring TPS — Tokens Per Second en 2026-10-29T17:35:49.310Z OpenAI expands misalignment monitoring across training, evaluation, internal use and deployment OpenAI, AI safety, misalignment, monitoring, Hugging Face, frontier models, evaluations