https://tpsreport.news/news/suno-speech-spoken-audio-background-music TPS — Tokens Per Second en 2026-10-02T20:20:45.414Z Suno adds Speech beta: spoken text and matching background music generated as a single audio track Suno, AI audio, speech generation, AI music, text-to-speech, copyright, product update https://tpsreport.news/news/cloudflare-clef-decision-models-64k-context-apache-2 TPS — Tokens Per Second en 2026-10-02T18:36:05.317Z Cloudflare releases Clef decision models, claims 39 ms median latency vs. 524 ms for TypeSafe's Jev Cloudflare, Clef, Clef-flash, decision models, AI agents, TypeSafe AI, Jev, Qwen https://tpsreport.news/news/gemini-app-map-tool-at-symbol-replaces-slash-skills TPS — Tokens Per Second en 2026-10-02T18:05:56.462Z Gemini app adds Map tool for location prompts on Android and iOS; '@' set to replace '/' for skills Gemini, Google, Gemini app, Map tool, skills, Connectors, Android, iOS https://tpsreport.news/news/openai-gpt-6-model-guide-startups-reasoning-effort TPS — Tokens Per Second en 2026-10-02T16:21:04.767Z OpenAI publishes startup guide for GPT-6 family covering model choice, reasoning effort and tool coordination OpenAI, GPT-6, developer guide, reasoning effort, tool use, startups, production workflows https://tpsreport.news/news/openai-gpt-6-startup-model-guide TPS — Tokens Per Second en 2026-10-02T16:20:43.399Z OpenAI publishes startup guide to choosing and deploying GPT-6 models, with reasoning-effort tuning OpenAI, GPT-6, reasoning effort, developer guide, agents, tool use, startups, prompt engineering https://tpsreport.news/news/claude-desktop-web-search-amazon-bedrock-agentcore TPS — Tokens Per Second en 2026-10-02T16:05:48.293Z AWS adds managed Web Search to Claude Desktop via Bedrock AgentCore Gateway in three Regions Amazon Bedrock, AgentCore, Claude Desktop, MCP, Web Search, AWS, Anthropic, enterprise AI https://tpsreport.news/news/ai2-astabrief-8b-open-source-report-generation TPS — Tokens Per Second en 2026-10-02T15:36:13.756Z Ai2 open-sources AstaBrief 8B, a Qwen3-8B report model it says runs 3.5x faster than Claude in Asta Ai2, AstaBrief, Asta, open-source, Qwen3-8B, scientific-research, report-generation, DPO https://tpsreport.news/news/gemini-4-argon-third-artificial-analysis-cyber-rollout TPS — Tokens Per Second en 2026-10-02T15:35:47.354Z Google's Gemini 4 Argon ranks third on Artificial Analysis index, rolls out first to cyber partners Google, Gemini 4, Gemini 4 Argon, Google DeepMind, Artificial Analysis, Claude Opus 5.5, Claude Sonnet 5.5, cybersecurity https://tpsreport.news/news/ling-3-1-flash-inclusionai-560b-moe-262k-context TPS — Tokens Per Second en 2026-10-02T14:50:49.275Z inclusionAI releases Ling 3.1 Flash: 560B MoE, 25B active, 262K context, free on OpenRouter Ling 3.1 Flash, inclusionAI, mixture-of-experts, MoE, hybrid reasoning, OpenRouter, 262K context, NovitaAI https://tpsreport.news/news/mercor-apex-accounting-benchmark-ai-vs-cpas TPS — Tokens Per Second en 2026-10-02T11:35:44.937Z Mercor study: AI models beat 12 licensed CPAs on simplified accounting tasks, but top score on full APEX benchmark is 61 Mercor, APEX Accounting, benchmark, Claude Opus 5.5, GPT-6 Astra, Fable 5.1, accounting, AI evaluation https://tpsreport.news/news/suno-speech-public-beta-ai-voice-music TPS — Tokens Per Second en 2026-10-02T09:50:41.638Z Suno launches Speech public beta: AI voiceovers with background music, up to about eight minutes Suno, Speech, text-to-speech, AI audio, AI music, voice generation, public beta, ElevenLabs https://tpsreport.news/news/mai-transcribe-2-streaming-60-languages-100ms TPS — Tokens Per Second en 2026-10-02T09:35:50.928Z Microsoft's MAI-Transcribe-2-Streaming returns first results in ~100 ms across 60 languages Microsoft, MAI, speech-to-text, text-to-speech, voice agents, transcription, voice cloning, Microsoft Foundry https://tpsreport.news/news/ramp-ai-index-spending-falls-usage-rises-anthropic-openai TPS — Tokens Per Second en 2026-10-02T09:20:44.802Z Ramp AI Index: US business AI spending falls while usage rises about 50% from July peak Ramp AI Index, AI spending, Anthropic, OpenAI, API pricing, enterprise AI, market share, open-source models https://tpsreport.news/news/pi-1-0-pi-durable-typescript-agent-harness-earendil TPS — Tokens Per Second en 2026-10-02T06:50:47.968Z Pi 1.0 agent harness goes stable; Pi Durable ports it to TypeScript with crash-resumable state Pi, Earendil, AI agents, agent harness, TypeScript, MCP, durable execution, developer tools https://tpsreport.news/news/cline-cli-v3-0-68-agent-teams-fix-model-catalog TPS — Tokens Per Second en 2026-10-02T04:50:42.795Z Cline CLI v3.0.68 fixes agent-team slowdown that could grow teams.db to gigabytes Cline, CLI, agent teams, coding agents, model catalog, release notes, DigitalOcean, Nvidia https://tpsreport.news/news/gpt-6-astra-ultrafast-nvidia-blackwell-8x-faster TPS — Tokens Per Second en 2026-10-01T23:50:43.366Z GPT-6 Astra Ultrafast Claims Up to 8x Faster Token Generation on NVIDIA Blackwell GPUs OpenAI, GPT-6, GPT-6 Astra, NVIDIA, Blackwell, inference, Codex, ChatGPT Work https://tpsreport.news/news/google-gemini-live-guided-vision-android-accessibility TPS — Tokens Per Second en 2026-10-01T20:05:48.066Z Google launches Guided Vision in Gemini Live, giving real-time audio descriptions through the Android camera Google, Gemini, Gemini Live, Guided Vision, accessibility, Android, TalkBack, multimodal https://tpsreport.news/news/cloudflare-clef-27b-decision-model TPS — Tokens Per Second en 2026-10-01T19:36:51.459Z Cloudflare releases Clef, a 27B Apache-2.0 model that outputs decision probabilities instead of text Cloudflare, Clef, Qwen3.8, multimodal, open-weights, Apache-2.0, decision-models, Hugging Face https://tpsreport.news/news/chatgpt-virtual-try-on-favorites-images-2-5 TPS — Tokens Per Second en 2026-10-01T19:36:16.656Z ChatGPT adds virtual try-on and Favorites in global shopping update built on Images 2.5 OpenAI, ChatGPT, ChatGPT Images 2.5, virtual try-on, shopping, e-commerce, image generation, product update https://tpsreport.news/news/gemini-4-argon-pricing-limited-access TPS — Tokens Per Second en 2026-10-01T19:35:58.974Z Google unveils Gemini 4 Argon at $2/$10 per 1M tokens, but access is limited to select users Gemini 4, Gemini 4 Argon, Google, Spark, personal agents, pricing, OpenAI, Meta Muse https://tpsreport.news/news/grok-for-intune-enterprise-ios-app-spacexai TPS — Tokens Per Second en 2026-10-01T18:35:42.979Z SpaceXAI launches Grok for Intune, a separate iOS app with Microsoft Intune app protection support Grok, SpaceXAI, xAI, Microsoft Intune, enterprise, iOS, mobile app, MAM https://tpsreport.news/news/tavus-griffin-human-interaction-model-48-percent-mistaken-for-human TPS — Tokens Per Second en 2026-10-01T18:20:47.081Z Tavus says 48% of testers mistook its Griffin video AI for a real person on a one-minute call Tavus, Griffin, Human Interaction Model, AI video avatar, real-time video, conversational AI, Nvidia, synthetic media https://tpsreport.news/news/opus-5-5-this-matters-graphite-ai-writing-tells TPS — Tokens Per Second en 2026-10-01T18:05:50.689Z Graphite: Opus 5.5 uses 'this matters' 116x more than humans as AI writing tells persist Claude Opus 5.5, Anthropic, OpenAI, Astra, Gemini 3.1 Pro, Graphite, AI writing detection, LLM style https://tpsreport.news/news/ideogram-4-5-localized-image-editing-native-2k TPS — Tokens Per Second en 2026-10-01T17:50:50.770Z Ideogram 4.5 launches with native 2K output and four tiers from $0.008 to $0.22 per image Ideogram, Ideogram 4.5, image generation, image editing, API, open weights, GPT-Image 2.5, Nano Banana https://tpsreport.news/news/amazon-strands-decider-2b-open-source-decision-model TPS — Tokens Per Second en 2026-10-01T17:05:52.724Z Amazon open-sources Strands Decider 2B, a small decision model built on a Qwen3.5-2B base Amazon, AWS, Strands Labs, decision models, open source, small models, AI agents, TypeSafe https://tpsreport.news/news/ambient-agents-amazon-bedrock-agentcore-ask-human TPS — Tokens Per Second en 2026-10-01T16:50:51.352Z AWS details ambient agent pattern on Bedrock AgentCore: S3 events trigger jobs, one ask_human tool pauses for approval Amazon Bedrock AgentCore, ambient agents, human-in-the-loop, AWS, agent infrastructure, serverless, Claude Sonnet 4.5, event-driven https://tpsreport.news/news/uniopen-amazon-nova-2-lite-lora-retail-moderation TPS — Tokens Per Second en 2026-10-01T15:50:45.682Z uniopen lifts Amazon Nova 2 Lite moderation F1 from 0.585 to 0.855 using LoRA fine-tuning on SageMaker AI Amazon Nova, Nova 2 Lite, fine-tuning, LoRA, SageMaker AI, content moderation, uniopen, AWS https://tpsreport.news/news/olmo-core-3-open-moe-training-trillion-parameters TPS — Tokens Per Second en 2026-10-01T15:20:50.975Z Ai2 releases Olmo-core 3, an open MoE training stack benchmarked at 1.2T parameters Olmo, Ai2, mixture-of-experts, MoE, training infrastructure, open source, NVIDIA B300, MXFP8 https://tpsreport.news/news/openai-dots-agent-gpt-6-astra-pro-plan TPS — Tokens Per Second en 2026-10-01T14:50:48.725Z OpenAI launches Dots agent on GPT-6 Astra, limited to $20/month Pro tier and above OpenAI, Dots, GPT-6 Astra, AI agents, Meta, Muse, DevDay 2026, enterprise AI https://tpsreport.news/news/claude-for-government-civilian-agencies-fedramp-high TPS — Tokens Per Second en 2026-10-01T14:35:44.083Z Anthropic launches Claude for Government for US civilian agencies in FedRAMP High environment Anthropic, Claude, Claude for Government, FedRAMP, Pentagon, US government, AI policy, enterprise AI https://tpsreport.news/news/unbiased-pareto-26-10-preview-1m-context TPS — Tokens Per Second en 2026-10-01T14:20:52.175Z Unbiased releases Pareto 26.10 Preview: 1M context, $0.80/$3.20 per 1M tokens on OpenRouter Pareto, Unbiased, OpenRouter, multimodal, 1M context, agentic, coding, model release https://tpsreport.news/news/openai-reasoning-extraction-campaign-azure-gpt-6-astra TPS — Tokens Per Second en 2026-10-01T12:20:46.364Z OpenAI says it shut down a 15,000-account campaign to extract model reasoning; the attack still worked on Azure OpenAI, distillation, model security, Azure, GPT-6 Astra, Moonshot AI, Anthropic, reasoning models https://tpsreport.news/news/gemini-4-argon-launch-pricing-benchmarks TPS — Tokens Per Second en 2026-10-01T09:35:52.964Z Google launches Gemini 4 Argon at $2/$10 per 1M tokens, rolling out first to Fairwind cybersecurity partners Gemini 4, Gemini 4 Argon, Google, Google DeepMind, GPT-6 Astra, Artificial Analysis, cybersecurity, API pricing https://tpsreport.news/news/gemini-4-argon-google-cybersecurity-model TPS — Tokens Per Second en 2026-09-30T23:50:50.107Z Google Releases Gemini 4 Argon, Positions It as Most Powerful Model Yet With Cybersecurity Focus Gemini, Google, Google DeepMind, cybersecurity, AI model release, Fairwind Program