https://tpsreport.news/news/openai-dots-task-router-gpt-6-1-terra-grok-bot-compute-costs TPS: Tokens Per Second en 2026-11-03T20:21:01.852Z OpenAI adds task router to Dots to cut compute as Musk calls Grok Bot costs 'unsustainably high' OpenAI, Dots, GPT-6.1 Terra, GPT-6 Astra, Grok Bot, xAI, AI agents, compute https://tpsreport.news/news/apple-siri-ai-app-skills-ios-27-3-developers TPS: Tokens Per Second en 2026-11-03T19:35:54.813Z Apple opens Siri AI to all third-party apps with 'app skills' in iOS 27.3 developer beta Apple, Siri AI, iOS 27.3, macOS 27.3, app skills, developers, App Intents https://tpsreport.news/news/openai-astra-physical-engineering-mazda-chip-ganassi TPS: Tokens Per Second en 2026-11-03T19:20:51.478Z OpenAI Unveils Astra-Designed Test Device, Expands Chip Ganassi Racing Work, Starts Mazda Research OpenAI, Astra, engineering, Mazda, Chip Ganassi Racing, vehicle design https://tpsreport.news/news/google-ple-85-task-engineering-benchmark-agent-competition TPS: Tokens Per Second en 2026-11-03T17:50:41.198Z Google releases 29 public tasks from 85-task engineering benchmark PLE and opens agent competition Google, benchmark, PLE, Mercor, AI agents, engineering, leaderboard https://tpsreport.news/news/statebench-ai2-long-context-state-tracking-benchmark TPS: Tokens Per Second en 2026-11-03T17:36:08.741Z Ai2's StateBench: Frontier models fall to 13% or lower on hardest long-input state tracking StateBench, Ai2, benchmark, long-context, state-tracking, Gated DeltaNet, Olmo Hybrid, transformers https://tpsreport.news/news/google-gemini-k12-students-off-by-default-schools-opt-in TPS: Tokens Per Second en 2026-11-03T17:05:55.717Z Google turns off Gemini for K-12 student accounts until schools opt in Google, Gemini, Google Classroom, Guided Learning, education, K-12, AI Search https://tpsreport.news/news/ibm-palette-20b-model-slide-decks-design-language TPS: Tokens Per Second en 2026-11-03T16:51:20.507Z IBM Research fine-tunes a 20B model to build slide decks in the IBM Design Language IBM, Palette, LoRA, fine-tuning, synthetic data, PresentBench, slide generation, open-weight https://tpsreport.news/news/github-copilot-jetbrains-air-acp TPS: Tokens Per Second en 2026-11-03T16:20:35.101Z GitHub Copilot is now available in JetBrains Air via the Agent Client Protocol GitHub Copilot, JetBrains Air, Agent Client Protocol, ACP, coding agents, GitHub, JetBrains https://tpsreport.news/news/cuga-hermes-deep-agents-appworld-glm-5-3-flash TPS: Tokens Per Second en 2026-11-03T16:05:59.500Z IBM's CUGA harness passes 81.1% of AppWorld tasks on GLM-5.3-flash; Hermes 46.7%, Deep Agents 30.0% CUGA, IBM, AppWorld, agent harness, GLM-5.3-flash, Hermes, Deep Agents, benchmark https://tpsreport.news/news/snowflake-coco-cowork-always-on-bots-private-preview TPS: Tokens Per Second en 2026-11-03T15:50:47.269Z Snowflake announces always-on CoCo and CoWork Bots that work in the background, coming soon to private preview Snowflake, CoCo, CoWork, AI agents, background agents, private preview, data engineering https://tpsreport.news/news/cogent-vr-1-1-livecvebench-57-9-percent TPS: Tokens Per Second en 2026-11-03T15:20:55.718Z Cogent's VR-1.1 Passes Repairs on 57.9% of LiveCVEBench Tasks, Versus 45.3% for Claude Code and Codex Cogent Security, VR-1.1, LiveCVEBench, vulnerability repair, Claude Code, Codex, benchmark https://tpsreport.news/news/gemini-4-carbon-intelligence-index-v5-59 TPS: Tokens Per Second en 2026-11-03T13:21:02.757Z Google's Gemini 4 Carbon scores 59 on Artificial Analysis Intelligence Index V5, one point ahead of Claude Opus 5.5 Gemini 4 Carbon, Google, Artificial Analysis, benchmarks, hallucination, Claude Opus 5.5, Fairwind, AA-Omniscience https://tpsreport.news/news/anthropic-flashprove-cryptographic-inference-proofs TPS: Tokens Per Second en 2026-11-03T05:05:38.141Z Anthropic releases FlashProve, open-source cryptographic proofs for AI inference, at 110 tokens per second Anthropic, FlashProve, cryptographic proofs, AI inference, verification, Llama 4 Maverick, open source https://tpsreport.news/news/zyphra-hierarchical-attention-zha-subquadratic-attention TPS: Tokens Per Second en 2026-11-03T03:05:41.381Z Zyphra introduces ZHA, tree-indexed sparse attention retrofitted onto frozen LLMs with 4B tokens of distillation Zyphra, ZHA, sparse attention, subquadratic attention, long context, distillation, research https://tpsreport.news/news/alibaba-cloud-model-studio-memory-personal-agents TPS: Tokens Per Second en 2026-11-03T02:35:50.409Z Alibaba Cloud Launches Memory Service on Model Studio, Claims Top Scores on HaluMem and PersonaMem-v2 Alibaba Cloud, Model Studio, agent memory, personal agents, MCP, HaluMem, PersonaMem-v2, Qwen https://tpsreport.news/news/tencent-hunyuan-hyra-2-0-research-agent-builds-scientific-tools TPS: Tokens Per Second en 2026-11-03T02:20:51.343Z Tencent Hunyuan releases Hyra 2.0, a research agent that builds reusable tools, under Apache 2.0 Tencent, Hunyuan, Hyra 2.0, research agent, protein folding, force fields, quantum error correction, open source https://tpsreport.news/news/qodo-codex-context-aware-code-review TPS: Tokens Per Second en 2026-11-02T22:35:55.099Z Qodo brings company knowledge and coding standards into Codex for pre-PR code review Qodo, Codex, OpenAI, code review, developer tools, AI coding https://tpsreport.news/news/openai-lifeval-everyday-tasks-evaluation TPS: Tokens Per Second en 2026-11-02T22:05:41.922Z OpenAI publishes LifeVal, an evaluation built partly from a 3,995-respondent survey to test AI on everyday tasks OpenAI, LifeVal, benchmark, evaluation, AI agents, everyday tasks https://tpsreport.news/news/sierra-agent-studio-ghostwriter-redesign TPS: Tokens Per Second en 2026-11-02T20:50:40.891Z Sierra rebuilds Agent Studio around Ghostwriter, its agent-building agent Sierra, Ghostwriter, Agent Studio, AI agents, product update https://tpsreport.news/news/gpt-6-astra-weight-space-programming-qwen3-5-0-8b-math-500 TPS: Tokens Per Second en 2026-11-02T20:20:49.185Z GPT-6 Astra edits 0.016% of Qwen3.5-0.8B weights, lifting MATH-500 from 7.9% to 48.1% GPT-6 Astra, Weight-Space Programming, Qwen3.5-0.8B, MIT CSAIL, interpretability, MATH-500, IFEval, research https://tpsreport.news/news/anthropic-message-threads-public-beta-messages-api TPS: Tokens Per Second en 2026-11-02T19:36:02.422Z Anthropic launches message threads in public beta, letting Messages API apps send only new messages Anthropic, Claude Platform, Messages API, message threads, Claude Code, AI agents, public beta https://tpsreport.news/news/deep-principle-hydrogen-1-0-753b-open-weight-scientific-model TPS: Tokens Per Second en 2026-11-02T19:05:50.029Z Deep Principle releases Hydrogen-1.0, a 753B open-weight scientific model under the MIT License Hydrogen-1.0, Deep Principle, open-weight, scientific AI, GLM-5.3, MIT License, chemistry, materials science https://tpsreport.news/news/google-genie-4-world-models-robotics-ai-testing TPS: Tokens Per Second en 2026-11-02T17:50:52.438Z Google Genie 4 points to new uses for world models in robotics and AI testing Google, Genie 4, world models, robotics, AI testing https://tpsreport.news/news/deepseek-harness-creator-mode-custom-plugins TPS: Tokens Per Second en 2026-11-02T17:50:36.624Z DeepSeek adds "creator mode" to DeepSeek Harness for building custom plugins DeepSeek, DeepSeek Harness, creator mode, plugins, AI agents, product update https://tpsreport.news/news/perplexity-brain-memory-architecture-agent-api-testing TPS: Tokens Per Second en 2026-11-02T16:35:55.411Z Perplexity details Brain, the self-improving memory system in Perplexity Computer, and Agent API memory tests Perplexity, Perplexity Computer, Brain, agent memory, Agent API, AI agents https://tpsreport.news/news/gemini-4-carbon-gdpx-benchmark-semianalysis TPS: Tokens Per Second en 2026-11-02T15:35:56.229Z Gemini 4 Carbon ranks second on SemiAnalysis GDPx at just over $16 per task, behind Claude Opus 5.5 Gemini 4 Carbon, Google, SemiAnalysis, GDPx, Claude Opus 5.5, GPT-6 Astra, benchmark, cost per task https://tpsreport.news/news/apple-online-dense-rl-planning-grounding-ios-ui-agent TPS: Tokens Per Second en 2026-11-02T14:36:18.123Z Apple researchers train 8B UI agent with online RL on live iOS apps, tying Gemini-3.1-Pro Apple, reinforcement learning, UI navigation, iOS, PPO, GiGPPO, agents, vision-language models https://tpsreport.news/news/openai-merges-chatgpt-chat-work-tabs-desktop TPS: Tokens Per Second en 2026-11-02T14:35:54.982Z OpenAI to merge ChatGPT's Chat and Work tabs on desktop; ChatGPT and Codex will share one token allowance OpenAI, ChatGPT, Codex, ChatGPT desktop app, product update, token allowance https://tpsreport.news/news/sensenova-6-8-flash-sensetime-multimodal-agent TPS: Tokens Per Second en 2026-11-02T14:05:55.679Z SenseTime launches SenseNova 6.8 Flash, a multimodal agent model, free for a limited time SenseNova, SenseTime, agent, multimodal, benchmarks, long-term memory, China AI https://tpsreport.news/news/glm-5-3-flashx-legal-1m-context-zai TPS: Tokens Per Second en 2026-11-02T13:50:56.255Z Z.ai releases GLM-5.3-FlashX-Legal: 1M-token legal model priced at $0.37/$1.25 per 1M tokens Z.ai, GLM-5.3, GLM-Legal, legal AI, multimodal, 1M context, OpenRouter, MCP https://tpsreport.news/news/grok-bot-usage-limits-throttled-unsustainable-costs TPS: Tokens Per Second en 2026-11-02T13:35:52.640Z Musk says Grok Bot limits were cut because serving costs were 'unsustainably high' Grok Bot, Elon Musk, usage limits, X, AI agents, AI costs https://tpsreport.news/news/hirundo-lazarus-realign-chinese-open-weight-models TPS: Tokens Per Second en 2026-11-02T12:50:49.684Z Hirundo and Lazarus AI edit weights of Chinese open models to strip censorship for Western clients machine-unlearning, Hirundo, Lazarus AI, Qwen, DeepSeek, open-weight models, AI safety, censorship https://tpsreport.news/news/neurodx-sphota-240m-model-liquid-ai-benchmark TPS: Tokens Per Second en 2026-11-02T07:50:53.845Z NeuroDX's 240M-parameter Sphota scores 52.4% vs Liquid AI's 52.1% across about 11 benchmarks, company claims Sphota, NeuroDX, Manas AI, Liquid AI, IndiaAI Mission, small language models, benchmarks https://tpsreport.news/news/semianalysis-gdpx-benchmark-opus-5-5-12-percent TPS: Tokens Per Second en 2026-11-02T05:35:49.332Z SemiAnalysis launches GDPx agentic benchmark across 11 domains; best model, Opus 5.5, scores 12% GDPx, SemiAnalysis, benchmark, Opus 5.5, GPT-6 Astra, Gemini 4 Carbon, agentic, GDPval https://tpsreport.news/news/gpt-6-2-astra-gray-test-claude-mythos-5-5-spotted TPS: Tokens Per Second en 2026-11-02T05:20:55.122Z GPT-6.2-Astra enters gray test for some Pro users as Claude Mythos 5.5 appears in site code GPT-6.2-Astra, Claude Mythos 5.5, OpenAI, Anthropic, gray test, Hugging Face, leak https://tpsreport.news/news/deepseek-dsh-superagent-open-source-coding-agent-prototype TPS: Tokens Per Second en 2026-11-02T04:51:02.329Z DeepSeek's Harness team open-sources dsh-superagent, a research prototype coding agent driven by GitHub DeepSeek, DeepSeek Harness, dsh-superagent, coding agent, open source, GitHub, Feishu https://tpsreport.news/news/meituan-longcat-2-5-1-6t-multimodal-agentic-coding TPS: Tokens Per Second en 2026-11-02T02:35:57.281Z Meituan releases LongCat-2.5: 1.6T-parameter multimodal model, 88.4 on Terminal Bench v2.1 Meituan, LongCat-2.5, agentic coding, multimodal, Terminal Bench, W4A4, mixture of experts