Developers are secretly hoping OpenAI fails to ship this month
The New Stack 1 week ago 18
OpenAI’s Codex sprint reported its first shipment, a speed-focused improvement that reduces the chance subscribers get an immediate free usage reset.
Model Releases
219 summarised stories about Model Releases, each linking back to the original source. Browse all topics →
Key players
Monday, 5 October 2026
Reflection debuts Beam, an open-weight AI model to rival Chinese models at lower compute cost
TechCrunch 1 week ago 10 ● 6 sources
Reflection AI unveiled Beam, its first frontier open-weight AI model meant to match leading Chinese open models on reasoning while costing less compute. Beam has 501 billion total parameters with 23 billion active and uses a 1 million token context window. The release positions Reflection to compete with closed labs and other open models and to enable “AI factory” deployments where institutions build customized local systems using Beam’s weights and technical details.
Meta Rushed to Fix Muse ‘VM Escape' Vulnerability Immediately Before Launch
404 Media 1 week ago 27 ● 33 sources
Meta engineers found multiple vulnerabilities in Muse, a viral AI agent, shortly before launch that could have let a user escape Muse’s per-user virtual machine and reach Meta-sensitive services. The internal hardening effort began on August 27 and ran for a handful of weeks, ending about 11 days before Muse’s release on September 8. Meta responded by rushing fixes to reduce reachable surface area and constrain network/port destinations for Muse agents, while continuing security work post-launch.
Remocn Studio
Product Hunt 1 week ago 45
Remocn Studio launched today as a free, open-source macOS AI video editor built from users’ Claude Code, Codex, Copilot, or Grok outputs into a Remotion-style project. It lets exports go up to 4K. The app changes video editing by replacing a timeline with live, clickable element editing, chat-based notes, and MP4 export from those generated projects.
The Story of Qwen: Alibaba’s AI Models From 7B to 2.4T
MarkTechPost 1 week ago 22
Alibaba’s Qwen model line evolved from an invite-only Tongyi Qianwen enterprise chatbot beta into open-weight releases that reached 2.4 trillion parameters. The Qwen3.8 release open-sourced Qwen3.8-2.4T-A95B with 2.4 trillion parameters and 95B active parameters on August 12, 2025. As Qwen grew, Alibaba shifted from mostly restrictive licensing to broader openness for smaller models, while making larger “Plus/Max” tiers proprietary and moving some revenue-based models to a two-tier split.
OrgComputers
Product Hunt 1 week ago 32
OrgComputers launched as a shared workspace for AI agents that brings projects, context, and tools together for use with Claude Code, Cursor, and Codex. The listing shows 9 followers. It changes how teams organize and reuse their existing knowledge when running AI coding tools.
Can an Open Model Do Security Research? Cantina’s apex-flash-1 Solves 40 of 60 Held-Out Bug Tasks
MarkTechPost 1 week ago 9
Cantina Security and Yeta Labs released apex-flash-1, an open-weights model trained for vulnerability research. It solved 40 of 60 held-out bug tasks (66.7% pass@1) with an estimated $2.38 per run versus $74.68 for Claude Opus 5 High. The release makes a locally deployable security-research “worker” model available under the MIT license, with noted hardware requirements (BF16 needing about 640 GB GPU memory).