Z.
Ox Alpha, a new reasoning model from Z.AI, is emerging as a contender in the heavyweight reasoning category, targeting DeepSeek's efficiency and reasoning capabilities. Early data shows stability in c…
Ox Alpha, a new reasoning model from Z.AI, is emerging as a contender in the heavyweight reasoning category, targeting DeepSeek's efficiency and reasoning capabilities. Early data shows stability in c…
China's zAI confirmed Wednesday that its Ox Alpha model is a new iteration of the GLM series, officially named GLM(-5.3 Flash), and will release its weights tonight, according to Bloomberg News. The f…
The Ox Alpha LLM has been identified as GLM-5.3-Flash, a new model from Zhipu AI that introduces a hybrid attention architecture combining 34 Kimi Delta Attention layers and 11 Multi-head Latent Atten…
China's Z.AI has developed Ox Alpha, a stealth AI model that rivals DeepSeek, according to a report. The model's existence was revealed despite the website blocking access to the original article. Z.A…
Z.AI, the Beijing AI lab co-founded by Zhang Peng, confirmed to Bloomberg on August 26 that the anonymous Ox Alpha model is a new iteration of its GLM series and plans to release its weights that nigh…
Mark Pesce, a technology writer, reported that he used AI agents to optimize his local Qwen3.8-27B model, achieving a 20% speed improvement after GPT-5.6 Sol in Codex spent four hours tuning it. He al…
Multiverse Computing researchers published a method called Quantization-Aware Healing on August 25 that shrank OpenAI's open GPT-OSS model from 120 billion parameters to 60 billion with 4-bit memory, …
CTGT founder and CEO Cyril Gorlla found that the anonymous coding model Ox Alpha, which appeared on OpenRouter on August 20 without a named developer, heavily suppresses answers about seven topics tie…
Ox Alpha, a model that appeared on OpenRouter on August 20, is behaviorally fingerprinted as part of the GLM family, with an exact 11-of-11 tokenizer match to the GLM-5.x vocabulary. The model censors…
A developer ran a week of real open source work on Ox Alpha, a free mystery coding model that appeared on OpenRouter under the identifier stealth/ox-alpha with no company attribution. The model handle…
Ox Alpha, a new reasoning-first AI model with a million-token context window, has appeared without an identified creator, and an independent community benchmark of 10 real-world coding tasks shows it …
An anonymous AI model, Ox Alpha, processed 11.6 trillion tokens in three days on OpenRouter, dwarfing the previous record by 2.6 times, driven by coding agents exploiting its 1,048,576-token context w…
Ox Alpha, an anonymous stealth AI model available for free on the Open Code coding agent platform, features a 1 million token context window, multimodal support, and a zero data retention claim, scori…
Ox Alpha, a free stealth AI model available through Open Code, can be paired with the open-source design tool Open Design to generate exportable HTML/CSS interfaces, with the free period reportedly en…
Ox Alpha, a newly announced free and multimodal AI model capable of processing a million tokens and videos, has reportedly outperformed Anthropic's Claude Fable, one of the leading AI models, accordin…
A mysterious AI model named Ox Alpha, listed as a stealth model on OpenRouter, has impressed developers with its coding and agentic capabilities, including a 1,048,576-token context window. An enginee…
A developer has documented how to run the unconfirmed Ox Alpha model in Claude Code without using a Claude subscription, by scoping an OpenRouter configuration to a single project folder. The setup re…
OpenCode reported that users processed 26 trillion tokens through the anonymous AI model Ox Alpha during its first four days, reaching 327,000 unique users and 8,328,244 completed sessions, making it …
A mysterious AI model called Ox Alpha, which benchmarks suggest rivals top models from Anthropic and OpenAI, appeared on the OpenRouter marketplace last week, but its creator remains anonymous. The mo…
Anthropic's Claude Code team confirmed an unannounced test that remapped the 'high' effort setting to the numeric value 10, previously used for 'low', in Claude Code 2.1.237, with credits promised for…