OpenCode and openrouter are releasing a model called 'stealth/ox-alpha' for free.
The context size is 1M, and the output is said to be 131,072 tokens.
It supports image and video input.
There are also articles saying that it beat fable or gpt-5.6 in user benchmark results...
(Retrieved from the internet. Source is unclear.)
It's not inconceivable that it can handle actual tool calls or long tasks... People speculate that it might be a new version of z.ai's GLM or a model from xiaomi.
Wow... The speed is so fast that I'm losing my mind. Anyway, I have to use it hard while it's free...
▶ Original source: https://openrouter.ai/stealth/ox-alpha
▶ Original source: https://opencode.ai/ko/go
