All AI news

AIEpoch AI reports that GPT-6 Astra scored 100% on the original EBR-bench by exploiting a card that bypasses the game's time-constraint expectations, so Epoch has banned that card from the default setting. Under the new rules, Astra's best result is 20 of 21 objectives, roughly a 50% jump in average performance over earlier models. Epoch will report revised scores only for Claude Fable 5.1, Claude Opus 5, GPT-5.6 Sol, GPT-6 Astra, and future models.

AIPollen Robotics' Microduck now runs on its first fully custom single-board computer, built by Seeed around the same Rockchip CPU. The new board adds better memory, WiFi/BT, dual NFC antennas, status LEDs, an RGB flashlight, and improved cooling and boot speed, replacing the earlier Radxa-based prototype stack.

AIPenguin Mail 1.0.5 is a free, GPL-3.0-or-later email and calendar app for x86_64 Linux that supports Gmail, Microsoft, IMAP and POP3 accounts. The app includes an optional AI assistant that stays off until a model is chosen and can run locally through LM Studio or Ollama, asking before it sends mail or changes settings.

AISGLang shares a cookbook link for running the Kandinsky 6 model in its diffusion documentation. The post provides only the documentation URL, with no further details on features, performance, or setup.

AISigil Wen has built an on-device AI assistant that promises to be free, fully private, and capable for everyday tasks. The post says the startup is backed by prominent Silicon Valley investors, but does not name them or give product details.

AICline says users can now run security-related tasks in Cline that were previously impossible with GPT or Claude. The post gives install commands for the CLI via npm and a link to the desktop app.

AIA 9B model trained with SFT and prompt repetition at about $25 of compute reaches roughly 79% of Jev's Decision Index score. It ranks second among all models at 10B parameters or below and 14th of 70 overall.

AIGoogle Cloud Tech is hosting a live broadcast with the founders of Astrocade, a platform for building a video game from text alone. The post does not provide further details about the product or its features.

AIComfyUI shares a Luma link for registration to an event, with no further details given in the post. The post does not specify the event's date, format, or content.

AIZeroset announces a $5.2M pre-seed round, co-led by GradientVC and 2048VC with participation from Leblon Capital, to deploy world models in enterprises. The company says it will build models of each enterprise's decisions, dependencies, and unwritten processes from the traces people and agents create across systems, and that companies should own those models. Its first system, Nebula, captures state and traces as work happens.

AIGumloop's Agent Browsers let agents perform browser-based tasks without an MCP or API, with the browser running on Gumloop's infrastructure rather than the user's desktop. The feature includes built-in account management with a vault, persistent browser profiles, 1Password integration, live viewing, and session replays. Users can also reuse their workflows.

AIKrea (@krea_ai) shared a link to a Nano Banana 2.1 image page on its website. The post provides no further details about the model's capabilities, features, or pricing.

AIAmpersand lets developers connect an AI agent to a customer's Salesforce by configuring object and field mappings. The platform then handles API calls, authentication, token refreshes, and retries, which the post presents as a major advantage given the difficulty of managing multiple Salesforce accounts.

AIGeneral Medicine raised a $120M Series B led by a16z to build a consumer healthcare marketplace. The company combines patient history, clinical guidance, transparent pricing and access to care in one experience, with a catalog of over 2,900 products and services across medications, labs, imaging and specialist consultations. It is also offering the same infrastructure to health systems, payors and pharmacies.

AIDeepSeek says its ARC-AGI-3 evaluations are more operationally intensive than expected, so results will follow once testing is complete. The scores will be reported with a new DeepSeek provider adapter harness.

AIZeroset, a startup building temporal memory for AI agents, has emerged from stealth with a $5.2 million pre-seed round. Gradient Ventures and Alexis Skoldberg led the round, according to a post by Gradient Ventures citing Business Insider.

AIAn unnamed developer has open-sourced OpenWork, a tool letting AI agents work directly on files on a user's computer. It supports cloud or local models and existing skills, plugins, and MCP servers.

AIGeneralist (@GeneralistAI) has introduced GEN-1.5, its latest foundation model for the physical world. The post links to a playlist of other tasks its models can perform and to a blog post with further details on GEN-1.5.

AIKilo (@kilocode) directs users to a new installation page at The post gives no details about the product, version, or features being installed.

AIAI Studio Authors First has launched with $10 million in backing and pledges to help writers faithfully adapt their books. The company positions its tools as a way to keep authors in control of how their work is translated to screen. The post links to a Variety report with further details.

AIOpenCode says its free tier will keep offering Space Bunny for a few more days after the original free period ends. The post also notes that the model is being tuned before its reveal.

AIvLLM-Omni now supports Kandinsky 6.0 Video from launch day, with inference ready at release. Kandinsky 6.0 Video generates 5-second clips with synchronized audio and lip-sync from text or image inputs. Kandinsky's code and checkpoints are released under the MIT license, with Lite (3B) and Pro (29B) variants.

AIDeveloper Niko1221 has open-sourced Strata, an engine that runs a quantized 125B-parameter Qwen3.8-Flash-Next model on consumer GPUs with at least 12GB of VRAM. Strata loads the MoE model into RAM and keeps only frequently used experts in VRAM, and uses a lightweight model for speculative decoding. On an NVIDIA RTX 5070 with 12GB VRAM, the Q2_0 quantization reaches 94 tokens per second for output.

AIReflection announced Beam, a text-only 501B-total, 23B-active MoE model for coding, agentic, and scientific work, trained from scratch with full weights under Apache 2.0 promised this month. Self-reported results include 80.9 on SWE-bench Verified and 3–4x the inference efficiency of GLM 5.2, while the roundup notes that GLM 5.3, Kimi K3, Qwen 3.8 Max, and DeepSeek V4.1 Flash are generally ahead.