--- title: "Consonance · CARD NO. 2026-09-27" description: "Xiaomi MiMo-V2.6 arrives on OpenRouter and vLLM — A radar over AI and agents. Every morning, and only what several independent sources are saying at once." canonical: "https://radar.consonance.fyi/en/jour/2026-09-27" lang: "en" published: "2026-09-27" updated: "2026-09-27T17:24:55.794Z" --- # Consonance · CARD NO. 2026-09-27 > A radar over AI and agents. Every morning, and only what several independent sources are saying at once. ## The day in brief _written by the model_ 1. **Xiaomi MiMo-V2.6 arrives on OpenRouter and vLLM** — Xiaomi's MiMo-V2.6 model series with a 1M-token context window launched simultaneously on both OpenRouter and the vLLM framework. (→ 20, 23) 2. **OpenAI and Anthropic launch new model families** — Both OpenAI and Anthropic introduced new model lines accompanied by significantly reduced operational costs compared to their previous iterations. (→ 09, 12) 3. **Google tests TPU processors in space orbit** — Google deployed satellite prototypes carrying Tensor Processing Units to evaluate machine learning workloads under harsh space conditions. (→ 02, 17) ## Video of the day **Anthropic launches Claude Opus 5.5 with 40% lower operational costs** — Anthropic has introduced Claude Opus 5.5, the first model in the Claude 5.5 family, delivering performance comparable to Claude Fable 5.1 while costing 40% less to run than Opus 5. The model features improved clarity and token efficiency across all effort levels. [Watch the video](https://radar.consonance.fyi/media/2026-09-27/en.mp4?v=13040126) · 1:11 · GENERATED AUTOMATICALLY - 0:00 Opening - 0:08 Claude Opus 5.5 - 0:20 ChatGPT Voice Integration - 0:37 Claude Code Updates - 0:51 Lightning Weave: Improving the Accuracy-Efficiency Frontier - 1:04 Closing ## Being discussed _Subjects at least three independent accounts raised over the last seven days._ ### 01. [Claude Code updates cloud sessions and usage limits](https://radar.consonance.fyi/en/sujet/957.md) _agents · 13 independent accounts · 76 posts · 3 labs · 204,916 interactions_ Claude Code officially released cloud sessions allowing tasks to run while laptops are closed, and added a graceful stopping point when hitting the 5-hour limit. The platform also resumed charging for requests blocked by safeguards. claude code voices: @ArtificialAnlys @ClaudeDevs @GithubProjects @Hesamation @addyosmani @aiedge_ @alex_prompter @alliekmiller ### 02. [Google tests TPUs in space and publishes new DeepMind essays](https://radar.consonance.fyi/en/sujet/958.md) _system_design · 9 independent accounts · 38 posts · 6 labs · 127,035 interactions_ Google's Project Suncatcher is testing a TPU-powered satellite prototype aboard a SpaceX mission. Additionally, Google DeepMind published essays addressing misbehavior control in agent swarms and equitable AGI benefit distribution. google · google deepmind voices: @Google @JeffDean @ShaneLegg @c_valenzuelab @dair_ai @dotey @elonmusk @firstadopter ### 03. [Hugging Face reviews internet access for AI agents during evaluation](https://radar.consonance.fyi/en/sujet/959.md) _evaluation · 5 independent accounts · 37 posts · 3 articles · 5 labs · 118,206 interactions_ Hugging Face is conducting an extensive review of AI agents' internet access during training and evaluation, publishing ongoing transparency summaries. Meanwhile, new diarization models were released to track multiple speakers. hugging face voices: @AndrewCurran_ @ClementDelangue @Gradio @Hesamation @NVIDIAAI @OpenAI @_akhaliq @kimmonismus ### 04. [ChatGPT Voice integrates plugins and expands pricing tiers](https://radar.consonance.fyi/en/sujet/960.md) _multimodal · 15 independent accounts · 76 posts · 4 labs · 114,972 interactions_ ChatGPT Voice added support for plugins like email, calendar, and Slack across web and mobile platforms. The ecosystem also prepares tiered subscription plans scaling up to a $500 Pro Max tier. chatgpt voices: @AndrewBolis @CompleteSkeptic @Hesamation @MatthewBerman @OpenAI @SchmidhuberAI @ajambrosino @alex_prompter ### 05. [Google introduces Gemini 3.8 Flash TTS and Flash-Lite TTS](https://radar.consonance.fyi/en/sujet/963.md) _multimodal · 13 independent accounts · 72 posts · 1 articles · 4 labs · 153,931 interactions_ Google launched Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, featuring over 2,000 production-ready voices, support for 100 languages, voice design tools, and voice replication. gemini 3.8 · gemini · gemini 3.1 flash voices: @AndrewCurran_ @ArtificialAnlys @EMostaque @GeminiApp @GithubProjects @Google @GoogleAIStudio @GoogleDeepMind ### 06. [Microsoft announces major GitHub Copilot update with Autopilot and GPT-6 models](https://radar.consonance.fyi/en/sujet/964.md) _agents · 3 independent accounts · 15 posts · 2 labs · 59,038 interactions_ Microsoft has announced a major update to GitHub Copilot, introducing proactive Autopilot and two new models: GPT-6 Sol for interactive coding and lightweight GPT-6 Luna. The release also includes a new Copilot application featuring Home, Code, and Copilot modes. github copilot voices: @AndrewBolis @emollick @mustafasuleyman @rohanpaul_ai @testingcatalog @alexeheath @github @jacobandreou Same story: [GitHub Copilot introduces Autopilot and integrates GPT-6 family models](https://radar.consonance.fyi/en/sujet/965.md) (12 voices) ### 07. [DigitalOcean launches Managed Agents in public preview](https://radar.consonance.fyi/en/sujet/966.md) _agents · 11 independent accounts · 61 posts · 1 articles · 2 labs · 224,616 interactions_ DigitalOcean has released Managed Agents in public preview, enabling users to run Claude Code, Codex, or LangGraph agents in runtime environments that pause when idle. The service places agent tools behind a governed endpoint and supports over 75 open-source options. codex voices: @GithubProjects @OpenAIDevs @aiedge_ @alex_prompter @alliekmiller @dani_avila7 @derrickcchoi @dotey ### 08. [Grok 4.7 xHigh scores 58% on Artificial Analysis AA-Briefcase benchmark](https://radar.consonance.fyi/en/sujet/967.md) _evaluation · 7 independent accounts · 36 posts · 3 labs · 207,318 interactions_ Grok 4.7 xHigh achieved 58% on the Artificial Analysis AA-Briefcase benchmark, ranking just one point behind Claude Fable 5.1 Max at 59%. Additionally, the stealth model Pixel Canary launched on Cline, tying with GPT-6 Astra and outperforming Kimi K3 on the Next.js Agent Evals benchmark. kimi k3 · glm · llama · llama.cpp voices: @Hesamation @aiedge_ @alex_prompter @cline @dotey @elonmusk @kimmonismus @rohanpaul_ai Same story: [xAI releases Grok 4.7 with a score of 46 on the Artificial Analysis Intelligence Index](https://radar.consonance.fyi/en/sujet/970.md) (10 voices) ### 09. [OpenAI introduces GPT-6 Sol and GPT-6 Luna models](https://radar.consonance.fyi/en/sujet/968.md) _architecture · 19 independent accounts · 85 posts · 1 articles · 8 labs · 311,465 interactions_ OpenAI has launched GPT-6 Sol and GPT-6 Luna, bringing the core strengths of GPT-6 Astra into faster and more affordable models designed for scaled workloads. Both models launch with API prices 50% lower than GPT-5.6, accompanied by Tesseract, a video creative suite built for AI agents. gpt-6 · deepswe · artificial analysis · terminal-bench voices: @AndrewCurran_ @AravSrinivas @ArtificialAnlys @Hesamation @OpenAI @OpenAIDevs @OpenRouter @PromptLLM Same story: [OpenAI releases GPT-6 Sol and Luna with a 50% permanent API price cut](https://radar.consonance.fyi/en/sujet/983.md) (3 voices) ### 10. [Meta announces new features and integration partners for Muse at Connect](https://radar.consonance.fyi/en/sujet/969.md) _agents · 10 independent accounts · 81 posts · 1 articles · 2 labs · 136,383 interactions_ At Meta Connect, Meta announced strategic integration partnerships for its personal AI agent Muse with Shopify, Expedia, and PayPal. These collaborations enable users to streamline shopping, trip planning, and checkout processes through Muse across participating merchants worldwide. muse spark voices: @AIatMeta @AlexFinn @aiedge_ @alliekmiller @dani_avila7 @dotey @emollick @finkd ### 11. [Anthropic launches plugin submission portal and expands MCP integration](https://radar.consonance.fyi/en/sujet/973.md) _agents · 6 independent accounts · 32 posts · 1 articles · 2 labs · 42,829 interactions_ Anthropic has launched a new portal allowing developers to submit plugins, track reviews, and monitor usage for Claude. These plugins package the Model Context Protocol (MCP) along with custom skills, serving as the standard mechanism for building applications on Claude. model context protocol · microsoft research voices: @ClaudeDevs @alex_prompter @bcherny @c_valenzuelab @dair_ai @dani_avila7 @dotey @levelsio ### 12. [Anthropic launches Claude Opus 5.5 with 40% lower operational costs](https://radar.consonance.fyi/en/sujet/975.md) _architecture · 22 independent accounts · 76 posts · 7 labs · 666,134 interactions_ Anthropic has introduced Claude Opus 5.5, the first model in the Claude 5.5 family, delivering performance comparable to Claude Fable 5.1 while costing 40% less to run than Opus 5. The model features improved clarity and token efficiency across all effort levels. claude fable voices: @AlexFinn @AndrewCurran_ @AnthropicAI @AravSrinivas @ArtificialAnlys @CuiMao @Hesamation @OpenRouter ### 13. [Claude Sonnet 5.5 spotted in stealth testing with a 1M-token context window](https://radar.consonance.fyi/en/sujet/978.md) _architecture · 3 independent accounts · 27 posts · 3 labs · 40,867 interactions_ Claude Sonnet 5.5 is currently undergoing stealth testing featuring a 1-million-token context window and a 128K-token maximum output length. Pricing is set at $2 per 1M input tokens and $10 for output tokens, arriving as a direct response to rival model releases. claude sonnet · claude haiku voices: @ClaudeDevs @CuiMao @GithubProjects @alexalbert__ @dani_avila7 @dotey @haider1 @kimmonismus ### 14. [Anthropic speeds up Claude.ai performance by 3x in two weeks](https://radar.consonance.fyi/en/sujet/980.md) _inference · 3 independent accounts · 5 posts · 2 articles · 1 labs · 50,040 interactions_ Anthropic has increased the performance speed of claude.ai and its desktop application by 3x within two weeks. The development team shared specific prompts and techniques utilizing Claude to measure, debug, and improve execution speed. voices: @ClaudeDevs @amorriscode @bcherny @trq212 @edwinarbus ### 15. [Grok 4.7 launches with support from NVIDIA accelerated computing](https://radar.consonance.fyi/en/sujet/981.md) _architecture · 9 independent accounts · 48 posts · 3 labs · 203,499 interactions_ SpaceXAI has released Grok 4.7, described as its most capable model yet for coding and knowledge work. The development of this system was supported by NVIDIA accelerated computing infrastructure. nvidia · flash attention · quantization voices: @ArtificialAnlys @CuiMao @Hesamation @NVIDIAAI @SchmidhuberAI @alex_prompter @elonmusk @firstadopter ### 16. [Sakana AI appoints Jürgen Schmidhuber as Chief Scientific Advisor](https://radar.consonance.fyi/en/sujet/982.md) _multimodal · 3 independent accounts · 20 posts · 2 articles · 2 labs · 42,482 interactions_ Jürgen Schmidhuber, a pioneer in meta-learning, recursive self-improvement, and world models since the 1990s, has joined Sakana AI as Chief Scientific Advisor. Concurrently, the company introduced Agora-2, a next-generation multi-agent world model supporting real-time interaction for up to 20 humans and agents. world model · david ha voices: @SchmidhuberAI @_akhaliq @alex_prompter @haider1 @hardmaru @rohanpaul_ai @AmOptimistShow @Markus50726803 ### 17. [Google launches a satellite to test TPU performance in orbit](https://radar.consonance.fyi/en/sujet/987.md) _system_design · 3 independent accounts · 11 posts · 1 labs · 25,820 interactions_ Google has partnered with Planet to launch a prototype satellite carrying four Tensor Processing Units (TPUs) into orbit. The mission evaluates how the hardware withstands harsh space conditions to assess the feasibility of hosting machine learning in orbit. moonshot ai voices: @Google @Hesamation @teortaxesTex @PeterDiamandis @TheStalwart ### 18. [Apple releases a Qwen3.5-9B finetune converting documents into images](https://radar.consonance.fyi/en/sujet/990.md) _architecture · 7 independent accounts · 33 posts · 2 articles · 1 labs · 20,123 interactions_ Apple has released a model on Hugging Face built by finetuning Qwen3.5-9B to convert long documents into small page images to conserve tokens, subsequently retrieving full text only for the pages relevant to a user's query. qwen3.8 · qwen3 · context window · kv cache · gemma · mixture of experts voices: @AlexFinn @ArtificialAnlys @OpenRouter @PrismML @dair_ai @heyshrutimishra @hwchase17 @rohanpaul_ai ### 19. [Agent harnesses integrate System One models and Contrastive Language Model](https://radar.consonance.fyi/en/sujet/1001.md) _agents · 3 independent accounts · 13 posts · 1 articles · 1 labs · 6,895 interactions_ The agent building community is experimenting with System One models like Jev and the Contrastive Language Model within custom harnesses. These models serve as high-speed routers or verifiers to improve performance for agent tasks. dspy voices: @CompleteSkeptic @lateinteraction @typesafeai @DSPyOSS @MaximeRivest @dbreunig @isaacbmiller1 @omarsar0 ### 20. [Xiaomi MiMo-V2.6 and System One model Jev launch on OpenRouter](https://radar.consonance.fyi/en/sujet/1002.md) _inference · 4 independent accounts · 37 posts · 1 articles · 13,047 interactions_ OpenRouter added several new models, including Xiaomi's multimodal MiMo-V2.6 series with a 1M-token context window and the System One model Jev. Developers are testing these models for applications requiring fast response times. openrouter voices: @CompleteSkeptic @LangChain @OpenRouter @giffmana @heyshrutimishra @KonstantinPilz @alexatallah @chromatique ### 21. [Xiaomi MiMo-V2.6-Pro tops open weights model benchmarks](https://radar.consonance.fyi/en/sujet/1005.md) _evaluation · 4 independent accounts · 4 posts · 1 labs · 14,067 interactions_ MiMo-V2.6-Pro achieved the top spot on the Artificial Analysis Intelligence Index for open weights models. The model utilizes a classic Grouped Query Attention architecture and delivers competitive cost efficiency per task. voices: @ArtificialAnlys @ClementDelangue @rasbt @teortaxesTex ### 22. [New research explores distillation and reinforcement learning for agents](https://radar.consonance.fyi/en/sujet/1008.md) _training · 4 independent accounts · 15 posts · 1 articles · 1 labs · 5,866 interactions_ Researchers released a post-training approach that helps agents learn from real user sessions by imitating successful trajectories and correcting errors. Another study from Google evaluates the impact of distilling agent harnesses on task success rates. distillation voices: @AravSrinivas @natolambert @rasbt @teortaxesTex @MatthewParrott @baselabs @omarsar0 @perplexity_ai ### 23. [vLLM adds day-one support for Xiaomi MiMo-V2.6 and DiffusionGemma-Jev](https://radar.consonance.fyi/en/sujet/1009.md) _inference · 4 independent accounts · 21 posts · 2 labs · 7,171 interactions_ The vLLM framework deployed day-one support for the Xiaomi MiMo-V2.6 models and DiffusionGemma-Jev. Both the Pro and Flash variants provide long-context and multimodal processing capabilities on the serving platform. vllm · sglang voices: @AravSrinivas @GithubProjects @teortaxesTex @vllm_project @PyTorch @SemiAnalysis_ @XiaomiMiMo @googlegemma ### 24. [Tencent ARC Lab releases GameHorizon Suite for evaluating game agents](https://radar.consonance.fyi/en/sujet/1014.md) _evaluation · 3 independent accounts · 11 posts · 1 articles · 1 labs · 3,652 interactions_ Tencent ARC Lab introduced the GameHorizon Suite, a data and evaluation framework designed to measure AAA gameplay capabilities across multiple temporal horizons for vision-language models, GUI agents, and coding agents. tencent · ling-3.0 · openbmb · zhipu ai voices: @ClementDelangue @levelsio @rohanpaul_ai @teortaxesTex @G2AIStack @HuggingPapers @OnSoloAI @Thom_Wolf ### 25. [Anthropic hackathon winner open-sources complete Claude Code setup](https://radar.consonance.fyi/en/sujet/1021.md) _agents · 4 independent accounts · 7 posts · 2,152 interactions_ The winner of Anthropic's hackathon has open-sourced their complete Claude Code setup, including agent skills, plugins, and practical tips. Concurrently, new research introduces methods to evolve agent skills while achieving 40 to 70 percent lower token costs compared to frontier methods, alongside NVIDIA's work on compiling public agent skills into reinforcement learning environments. agent skills voices: @NVIDIAAI @aiedge_ @dair_ai @dani_avila7 @tom_doerr ### 26. [Alibaba announces infrastructure plans and next-generation Qwen models](https://radar.consonance.fyi/en/sujet/1028.md) _system_design · 3 independent accounts · 15 posts · 1,427 interactions_ Alibaba aims to reach 20 gigawatts of global data center capacity by 2032 to support its artificial superintelligence ambitions, alongside plans for a 5 to 10 trillion parameter Qwen model and proprietary AI chips. Additionally, the Qwen team released RecreationWorld on Hugging Face as a sandbox for hybrid computer-use agents to explore, implement, and visually verify builds. alibaba · alibaba qwen voices: @ArtificialAnlys @Hesamation @alex_prompter @teortaxesTex @AlibabaGroup @HuggingPapers @poezhao0605 @tphuang ## Worth reading closely Nothing yet. ## Hands-on _Repositories climbing on GitHub today, scored on the day's momentum, rank, adoption, freshness and project health._ - [paperclipai/paperclip](https://github.com/paperclipai/paperclip): An open-source management platform for backend developers to orchestrate and monitor multi-agent workflows in enterprise environments. · +2,527 stars today · ★ 89,110 · TypeScript - [vectorize-io/hindsight](https://github.com/vectorize-io/hindsight): A long-term memory layer for AI agents that extracts, persists, and learns from historical interactions. · +4,463 stars today · ★ 36,336 · Python - [debpalash/VoiceStudio](https://github.com/debpalash/VoiceStudio): This fully local voice cloning and audio processing suite suits backend engineers integrating speech capabilities into AI agents. · +3,060 stars today · ★ 39,304 · Python - [rohitg00/ai-engineering-from-scratch](https://github.com/rohitg00/ai-engineering-from-scratch): A hands-on guide for backend engineers to build AI applications and agents from scratch. · +848 stars today · ★ 59,007 · Python - [NVIDIA/Model-Optimizer](https://github.com/NVIDIA/Model-Optimizer): NVIDIA's official library for quantizing and optimizing deep learning models to maximize inference throughput. · +357 stars today · ★ 4,833 · Python - [zhaoxuya520/reverse-skill](https://github.com/zhaoxuya520/reverse-skill): An AI-powered toolchain router and knowledge base for coding agents like Claude Code and Cursor. · +361 stars today · ★ 38,152 · PowerShell - [openbao/openbao](https://github.com/openbao/openbao): An open-source secrets management solution to safely store and distribute API keys for backend services and AI pipelines. · +364 stars today · ★ 8,061 · Go - [dream-num/univer](https://github.com/dream-num/univer): An integrated spreadsheet, doc, and canvas runtime for engineers building agents that need programmatic Office document manipulation. · +920 stars today · ★ 20,015 · TypeScript - [tensorflow/tensorflow](https://github.com/tensorflow/tensorflow): The classic machine learning framework for training deep neural networks, geared toward ML engineers. · +46 stars today · ★ 200,507 · C++ - [microsoft/vscode](https://github.com/microsoft/vscode): The quintessential extensible code editor for software developers building applications and tooling. · +95 stars today · ★ 193,118 · TypeScript - [mobile-next/mobile-mcp](https://github.com/mobile-next/mobile-mcp): A Model Context Protocol server for engineers building mobile automation and device scraping agents. · +168 stars today · ★ 7,515 · TypeScript - [vercel-labs/scriptc](https://github.com/vercel-labs/scriptc): A TypeScript-to-native compiler that improves execution performance for backend developers building high-speed services. · +76 stars today · ★ 5,301 · TypeScript - [mvschwarz/openrig](https://github.com/mvschwarz/openrig): A multi-agent harness orchestrating Claude Code and Codex together, ideal for engineers building automated coding agents. · +114 stars today · ★ 780 · TypeScript - [llvm/llvm-project](https://github.com/llvm/llvm-project): A collection of modular compiler technologies for systems engineers building custom toolchains and optimizers. · +41 stars today · ★ 40,779 · LLVM - [vercel/next.js](https://github.com/vercel/next.js): The standard React framework for full-stack web developers building production applications. · +54 stars today · ★ 142,708 · JavaScript - [anthropics/claude-code-action](https://github.com/anthropics/claude-code-action): A GitHub Action integrating Claude Code into CI/CD pipelines for engineers automating code reviews. · +31 stars today · ★ 9,153 · TypeScript ## Next - [Previous day: 26 September](https://radar.consonance.fyi/en/jour/2026-09-26.md) - [Next day: 28 September](https://radar.consonance.fyi/en/jour/2026-09-28.md) - [archives](https://radar.consonance.fyi/en/archives.md) - [method](https://radar.consonance.fyi/en/methode.md) - [Consonance](https://radar.consonance.fyi/en/index.md) - Languages: [Français](https://radar.consonance.fyi/jour/2026-09-27.md) · [Tiếng Việt](https://radar.consonance.fyi/vi/jour/2026-09-27.md) - For machines: [llms.txt](https://radar.consonance.fyi/llms.txt) · [RSS](https://radar.consonance.fyi/en/rss.xml) · [sitemap.xml](https://radar.consonance.fyi/sitemap.xml) - HTML version: https://radar.consonance.fyi/en/jour/2026-09-27