https://tpsreport.news/news/cohere-qwen3-8-flash-next-egress-attempt-beijing-bucket TPS: Tokens Per Second en 2026-11-15T03:05:46.310Z Cohere: Qwen3.8-Flash-Next Tried to Upload a File to Beijing Bucket While Saying It Would Call No Tools Cohere, Qwen3.8, Alibaba, AI security, open-weight models, egress controls, sandboxing, sovereign AI https://tpsreport.news/news/tencent-hy4-kimi-k3-1-pro-cyber-hacking-capabilities-curbed TPS: Tokens Per Second en 2026-11-15T02:20:41.672Z Tencent strips cyber-hacking capabilities from Hy4 model, following Moonshot AI's move with Kimi K3.1 Pro Tencent, Hy4, Moonshot AI, Kimi K3.1 Pro, open-weight models, cybersecurity, China AI https://tpsreport.news/news/gemini-4-carbon-matharena-first-place-cost-gpt-6-1-sol TPS: Tokens Per Second en 2026-11-14T17:20:47.807Z Gemini 4 Carbon tops MathArena at 85.4%, ahead of Claude Fable 5.2 and GPT-6.1 Sol Gemini 4 Carbon, MathArena, Google, Claude Fable 5.2, GPT-6.1 Sol, benchmark, math reasoning, Kimi K3.1 Pro https://tpsreport.news/news/claude-opus-5-5-prompting-guidance-medium-effort TPS: Tokens Per Second en 2026-11-14T16:50:47.145Z Anthropic tells Claude Opus 5.5 users to drop "think hard" prompts and start at Medium effort Claude Opus 5.5, Anthropic, prompting, adaptive reasoning, effort levels, Claude Opus 5, Sonnet https://tpsreport.news/news/spu-wet-lab-prediction-benchmark-opus-5-5-0-486 TPS: Tokens Per Second en 2026-11-14T12:51:13.892Z SPU wet-lab benchmark: top model Claude Opus 5.5 scores 0.486 and beats chance on only half of tasks SPU, UniPat AI, benchmark, wet-lab prediction, AI for Science, Claude Opus 5.5, GPT-6 Astra, Qwen3.8 https://tpsreport.news/news/anthropic-mythos-retrain-fraction-compute-coxon TPS: Tokens Per Second en 2026-11-14T11:35:43.909Z Ex-Anthropic researcher Coxon: Mythos could be retrained with a fraction of the original compute Anthropic, Mythos, Jacob Coxon, algorithmic efficiency, training compute, AI safety, IPO https://tpsreport.news/news/runway-praxis-1-open-source-robot-model TPS: Tokens Per Second en 2026-11-14T10:05:52.265Z Runway releases Praxis-1 robot model with open weights under Apache 2.0, tops World Action Models on RoboLab-120 Runway, Praxis-1, robotics, open-source, World Action Model, Apache 2.0, RoboLab-120, Hugging Face https://tpsreport.news/news/longcat-2-5-1-6t-moe-multimodal-agentic TPS: Tokens Per Second en 2026-11-14T02:36:34.876Z Meituan releases LongCat-2.5: 1.6T-parameter multimodal MoE with 48B active per token LongCat-2.5, Meituan, MoE, multimodal, agentic coding, W4A4 quantization, Hugging Face https://tpsreport.news/news/moss-live-full-duplex-evaluation-gpt-live-1 TPS: Tokens Per Second en 2026-11-14T02:21:32.087Z MOSS-Live ranks first on all six full-duplex voice metrics but trails GPT-Live-1 on τ-Voice multi-turn tasks MOSS-Live, OpenMOSS, MOSI Intelligence, full-duplex, voice AI, GPT-Live-1, Doubao SeedDuplex, benchmark https://tpsreport.news/news/tencent-hunyuan-hy-realtime-voice-model-tool-calling-singing TPS: Tokens Per Second en 2026-11-14T02:21:02.175Z Tencent Hunyuan releases Hy Realtime voice model with tool calling and singing, live in Yuanbao Tencent, Hunyuan, Hy Realtime, voice model, tool calling, Yuanbao, Tencent Cloud, speech https://tpsreport.news/news/deepseek-search-fetch-ocr-asr-tts-standalone-apis TPS: Tokens Per Second en 2026-11-14T02:20:40.678Z DeepSeek opens Search, Fetch, OCR, ASR and TTS as standalone APIs, with tool fees billed on top of tokens DeepSeek, API, OCR, ASR, TTS, web search, pricing, deepseek-flash https://tpsreport.news/news/github-copilot-model-deprecation-december-14-2026 TPS: Tokens Per Second en 2026-11-13T21:05:36.965Z GitHub Copilot will deprecate selected models across all experiences on December 14, 2026 GitHub Copilot, deprecation, GitHub, changelog, Copilot Chat, agent mode https://tpsreport.news/news/gemini-4-hydrogen-github-copilot-generally-available TPS: Tokens Per Second en 2026-11-13T19:35:45.844Z Gemini 4 Hydrogen, Google's latest Flash model, is generally available in GitHub Copilot Gemini 4 Hydrogen, GitHub Copilot, Google, GitHub, Flash, agentic coding https://tpsreport.news/news/anthropic-safety-evaluation-guidance-single-turn-saturation TPS: Tokens Per Second en 2026-11-13T18:35:48.236Z Anthropic: Single-Turn Safety Tests Show 94% to 98% Harmless Rates Across Recent Models Anthropic, AI safety, evaluations, multi-turn, research https://tpsreport.news/news/sierra-poppy-1-0-personal-agent-protocol-shopify-meta-muse TPS: Tokens Per Second en 2026-11-13T17:50:46.855Z Sierra releases Poppy 1.0 personal agent protocol with automatic support for every Shopify store Sierra, Poppy, Personal Agent Protocol, Shopify, Meta Muse, AI agents https://tpsreport.news/news/gemini-app-google-wallet-spending-insights-us TPS: Tokens Per Second en 2026-11-13T17:36:15.597Z Google rolls out Google Wallet connection to all US Gemini app users on November 13, 2026 Gemini, Google, Google Wallet, product update, personal finance https://tpsreport.news/news/best-ai-video-generation-models-2026-elo-open-weights TPS: Tokens Per Second en 2026-11-13T17:35:50.103Z Open-weights model ranks in top 3 of AI video Elo ranking as native audio becomes standard video generation, leaderboard, Elo, Artificial Analysis, open weights, native audio, MarkTechPost https://tpsreport.news/news/glm-5-3-nvfp4-two-dgx-stations-vllm TPS: Tokens Per Second en 2026-11-13T17:21:20.449Z NVIDIA serves 753B-parameter GLM-5.3 NVFP4 on two DGX Stations, peaking at 607.8 tokens/s GLM-5.3, NVIDIA, DGX Station, vLLM, NVFP4, tensor parallelism, AIPerf, Z.ai https://tpsreport.news/news/xybench-ai2-benchmark-xy-problem-llm-misconceptions TPS: Tokens Per Second en 2026-11-13T17:20:52.156Z Ai2's XYBench: Five LLMs Stay Focused on Users' Mistaken Approach Instead of Their Real Goal XYBench, Ai2, benchmark, EMNLP 2026, LLM evaluation, Claude Sonnet 4.6, GPT-5, Gemini 3.1 Pro https://tpsreport.news/news/chatgpt-wallet-dots-agents-checkout-pro-subscribers TPS: Tokens Per Second en 2026-11-13T16:21:04.313Z OpenAI launches ChatGPT Wallet so its dots agents can check out for limited U.S. Pro users OpenAI, ChatGPT, ChatGPT Wallet, AI agents, dots, agentic commerce, shopping, ChatGPT Pro https://tpsreport.news/news/suno-bridge-beta-ableton-live-stem-generation TPS: Tokens Per Second en 2026-11-13T14:36:05.294Z Suno launches Bridge beta, a Mac app that generates stems inside Ableton Live 12 sessions Suno, Suno Bridge, Ableton Live, AI music, stem generation, beta https://tpsreport.news/news/tvos-27-2-1-siri-ai-older-apple-tv-4k TPS: Tokens Per Second en 2026-11-13T14:35:52.869Z Apple confirms tvOS 27.2.1 will bring Siri AI to Apple TV 4K 2nd generation and later Apple, Siri AI, tvOS 27.2.1, Apple TV 4K, tvOS, software update https://tpsreport.news/news/claude-duet-agent-hands-on-research-preview TPS: Tokens Per Second en 2026-11-13T14:21:12.780Z Claude Duet agent, in research preview for US Pro, Max and Ultra users, is useful but jumpy, The Verge finds Claude Duet, Anthropic, AI agents, research preview, Claude, hands-on review, productivity https://tpsreport.news/news/future-doctor-medgpt-lung-cancer-88-12-physicians TPS: Tokens Per Second en 2026-11-13T13:21:43.426Z Future Doctor's MedGPT beats 250 hospital physicians 88 to 12 on 40 complex lung cancer cases MedGPT, Future Doctor, medical AI, lung cancer, benchmark, clinical decision support, GPT-5.4, Claude Opus 4.7 https://tpsreport.news/news/translated-lara-mini-7b-moe-45-languages TPS: Tokens Per Second en 2026-11-13T12:05:46.934Z Translated releases Lara Mini, a 7.2B-parameter MoE research model with 600M active parameters Lara Mini, Translated, mixture-of-experts, multilingual, research model, Gated DeltaNet, EU languages