Anthropic released Claude Haiku 5.5, a small model priced at $0.10 per million input tokens that retains a 1M-token context window and scores 72.4% on OSWorld, according to MarkTechPost. The release p…
Anthropic released Haiku 5.5 on October 7, 2026, cutting per-token prices 90% for requests under 100,000 tokens to $0.10 per million input and $0.50 per million output, matching GPT-6 Luna's rate. Ind…
Anthropic released Claude Haiku 5.5, which jumps from 15.7 percent to 72.4 percent on the OSWorld computer use benchmark versus its predecessor while cutting token prices by up to 90 percent. A new to…
Researchers Shashank Kirtania and co-authors introduced BREW (Bootstrapping expeRientially-learned Environmental knoWledge), a framework that distills an LLM agent's past interaction trajectories into…
H Company released Holo4 on September 28, 2026, a series of open-weight agentic models in 27B dense and 35B-A3B Mixture of Experts configurations that interact with software through GUIs, code executi…
Developers at Mininglamp built Mano-AFK, an open-source autonomous execution framework that takes a natural language prompt through PRD generation, coding, local deployment, three-layer testing, and a…
A head-to-head comparison finds Anthropic's Claude Sonnet 4.5 outperforms OpenAI's GPT-5 on coding and agentic benchmarks, scoring 77.2% on SWE-bench Verified versus GPT-5's 74.9%, and 50.0% versus 43…
Vercel now supports running Harbor evaluations, including Terminal-Bench, SWE-bench, tau3-bench and OSWorld, on Vercel Sandbox, with each trial executing in its own isolated Firecracker microVM when u…
A research note argues that the AI release slowdown debate conflates release cadence with training cadence, contending that coding, math and cyber are being remade where RL data and feedback loops exi…
UC Berkeley researchers released CUA-Lite, an open platform unifying sandboxes, data, evaluation, and reinforcement learning for computer-use agents, with Lite.OSWorld reproducing the OSWorld task sui…
OpenAI released GPT-6 Astra on 3 September 2026, its most capable model to date, with agentic scores of 74.1% on DeepSWE v1.1 and 72.6% on OSWorld 2.0. The model is the first to reach OpenAI's 'Critic…
Qwen has released Qwen3.8-27B, a 27-billion-parameter multimodal language model with integrated vision capabilities, built on the Qwen3.5 architecture. The model supports native image and video unders…
Anthropic's Claude Mythos Preview, Fable 5, and Opus 4.8 top the OSWorld-Verified leaderboard with scores of 85.4%, 85.0%, and 83.4%, up from 12% in April 2024, surpassing the human baseline of 72%. H…
AGI Ranker, an independent AI leaderboard, released v2.0.0 of its AGI Score, causing every model's score to drop by 6 to 15 points due to a correction in how human-parity ceilings are applied. The rev…
Anthropic's Claude Opus 5, the first 5th-generation Opus model, launched on Amazon Bedrock on July 24, claiming to match Fable 5 intelligence across agentic coding, knowledge work, visual understandin…
Google shipped Gemini 3.6 Flash on July 21, 2026, offering 17% fewer output tokens than its predecessor and available immediately in GitHub Copilot for Pro, Pro+, Max, Business, and Enterprise users. …
A new framework called BREW (Bootstrapping expeRientially-learned Environmental knoWledge) enables LLM agents to retain and use experience-based knowledge, improving task success rates by 10-20% and r…
Two of China's largest AI model platforms, Doubao and Qwen, tightened restrictions on custom agents last week, delisting or throttling third-party agents and narrowing creation permissions. The develo…
Researchers have developed a failure-driven self-improvement loop for AI training that leverages failed trajectories to boost performance, challenging the traditional focus on successes. The approach …
Mininglamp Technology has open-sourced Octo, a collaboration platform designed to connect multiple AI agents within teams. The platform addresses the lack of infrastructure for agent-to-agent communic…