Cohere Labs releases ATE dataset: only 2.6% of 696K MCP tools can fully execute professional tasks
Cohere Labs aggregated seven public directories to build the Agentic Task Ecosystem dataset, containing 696,291 tools…
AI news monthly archive: September 2026 features 473 curated AI news items covering model releases, product updates, industry news, papers, tutorials and opinions, each with an AI summary, source credit and original link.
Cohere Labs aggregated seven public directories to build the Agentic Task Ecosystem dataset, containing 696,291 tools…
ARC Prize reports that OpenAI's GPT-6 Astra scored 62.7% on ARC-AGI-3 Semi-Private with the Standard harness ($26K…
Google employee Shir Meir Lador introduces harness engineering, which wraps an LLM in deterministic components…
OpenAI and CEO Sam Altman face 30 new lawsuits accusing them of providing substantial assistance and encouragement to…
Unsloth used MTP (Multi-Token Prediction) to speed up local inference for Qwen3.8-Flash-Next by about 1.3 to 1.7x with…
Cursor released Self-Hosted Machines, which moves tool execution for cloud agents to machines within an enterprise's…
Baseten published an analysis of the cost drivers of post-training open-source models, noting that active parameter…
LongCat-2.0 is now available for free trial in @cline! 🐱 [Quoting @cline]: LongCat-2.0 is currently free to use in…
PromptArmor analyzed the security and cost of Claude Code and Cowork based on 310,009 OTel telemetry events from a…
UU Remote launched a new version on September 2, focusing on terminal improvements and adding TUI rendering…
Alibaba's Qwen released Qwen3.8-Max-0902, which debuted at No. 1 on the overall Code Arena: WebDev leaderboard with…
Bloomberg reports that Nvidia is nearing an acquisition of Hugging Face for about $12.9 billion, with the total deal…
Cursor added support for self-hosted machines, where codebases, build artifacts, and secrets remain on machines within…
Meta released Muse Spark 1.3, with performance improvements on agent and coding tasks; the max reasoning version is…
Anthropic launched Claude Fable 5.1 and Claude Mythos 5.1, calling them its most advanced models for coding and…
Rohan Paul reviewed the safety findings in the Fable 5.1 system card: Anthropic says the model achieves the highest…
Google DeepMind launched agentic video understanding for Gemini 3.7 Flash, 3.6 Flash and 3.5 Flash-Lite. The models…
Google released Google Pics, an image creation and editing tool for Workspace, which will roll out over the coming…
Anthropic published a guide to building commerce agents, summarizing practices across architecture, latency and cost…
Anthropic released a commerce agent blueprint for e-commerce, travel, telecom and ticketing platforms, offering a…
A Tessl designer reviews how brand and design rules were encoded into AI workflows in the agent software factory…
Meta released Muse Spark 1.3, its fourth Muse Spark version in five months. The xhigh version scores 61 on the…
Google DeepMind released Gemini 3.8 Flash, which scores 59 on the Artificial Analysis Intelligence Index with high…
OpenAI announced that Astra reaches the Critical cybersecurity capability threshold under its Preparedness Framework,…
According to Reuters, power requests from ultra-large electricity users, mainly data centers, in the Midwest,…
Shanghai AI Lab has released InternLumina-U2, a 16B-parameter (1B activated) MoE diffusion language model that uses an…
Anthropic published new research, Training a Misaligned Reward Seeker, examining whether reward hacking can teach…
The Hugging Face WebAI team released the @huggingface/kernels library and 207 WebGPU kernels (Apache-2.0) hosted as…
Anthropic released Claude Fable 5.1 (claude-fable-5-1) for long-running agentic coding, knowledge work and research,…
Meta Superintelligence Labs released Muse Voice Transcribe, which it calls the first real-time audio perception model,…
Anthropic released Claude Fable 5.1 and Claude Mythos 5.1. The two are the same model with different levels of safety…
Dario Amodei published a long article arguing for advancing AI capabilities at a balanced pace so that safety and…
Artificial Analysis evaluated Claude Fable 5.1, which tops the Artificial Analysis Intelligence Index with 66 points,…
Page 12 of 12
Sponsored
Orange Relay (pro.nat6.net) is a premium AI API relay — an excellent transfer station with competitive pricing, reliable quality and broad model coverage including Claude, GPT and Gemini.
Visit Orange Relay