New this month
Added Aug 20, 2026
Ox Alpha is a stealth AI model that appeared on OpenRouter and OpenCode on 20 August 2026. Its creator is unidentified, and OpenRouter routes requests to an anonymous third-party provider. The model accepts text, images and video, outputs text, and supports a 1,048,576-token context window with up to 131,072 tokens of output. Independent serving-layer forensics published on 22 August 2026 point to Zhipu AI's GLM-5.x infrastructure as the leading theory — a Java stack trace naming Zhipu's internal API classes, matching error-code dialects, 30/30 tokenizer alignment with GLM-5.3, and identical video-encoder behaviour to GLM-5V-Turbo. Zhipu has not confirmed this.
Why: The combination of a one-million-token context window, multimodal inputs, free pricing during the preview, and rapid adoption by coding-agent builders makes it worth tracking even before its creator is known. Within a day of launch, coding agents had pushed billions of tokens through it, suggesting real production interest rather than curiosity traffic. Preliminary independent DeepSWE testing also places it ahead of Claude Fable 5 and GPT-5.6 Sol on a small task subset.
New this month
Added Aug 20, 2026
NanoClaw is an open-source personal AI agent runtime built by Gavriel Cohen as a smaller, auditable alternative to OpenClaw. Each agent session runs in its own Docker container with scoped permissions and self-destructs when the task ends. In August 2026 it added a Slack integration that lets you provision persistent AI agent teams and colleagues from a single message, running on customer infrastructure.
Why: The agent landscape is polarised between all-in-one platforms with huge codebases and small, custom rigs. NanoClaw occupies the small, auditable end: roughly 500 lines of TypeScript, container isolation by default, and a fork-and-own model that makes the agent's capabilities explicit rather than hidden behind plugins.