Introduction of Ox Alpha Model: A New Player in Coding
OpenRouter has unveiled a new model, Ox Alpha, which shows impressive results in coding tasks. The model features a 1M token context and is available via API.
OpenRouter has introduced a mysterious model, Ox Alpha, which surpasses Sol and Fable in coding tasks. No one knows who the developer is, and the model is listed as stealth. Here’s what is known:
- 1M token context.
- Multimodal.
Ox Alpha has turned out to be GLM-5.3-Flash: the official release has occurred
Natively multimodal, optimized for efficient coding and long agent tasks. 1M context.
In terms of metrics, it ranks approximately at Opus 4.8, and in some areas, even better. It achieved an excellent score on the Automation Bench: 48.8 compared to 41 for Opus 4.8 and 37.2 for GPT-5.6 Terra. Moreover, it is a small model: 320B-A18B.
The model is available to all users of the GLM Coding Plan + in the API.
Current pricing: $0.075 per million input tokens, $0.25 per million output tokens (this includes a 50% discount on Open Router).
P.S. An interesting architectural detail is the hybrid attention. Z.ai combined sparse + linear attention to maintain accuracy on long contexts without inflating computational costs.
Why it matters
AnalysisThis model could significantly change coding approaches due to its multimodality and efficiency. It offers new opportunities for developers and researchers in the field of artificial intelligence.
Discuss in community
Share your questions and insights with developers
Related in Ecosystem
Practical guides, core terminology, and agent skills
GPT Image Generative Models Prompting Guide
OpenAI's gpt-image generative models are designed to produce production-quality visuals and fully controllable creative workflows. They are suitable for both professional design ta...
Claude Sonnet 4.6 vs GPT-5.5 vs Gemini 3.1 Pro: 2026 model comparison
Every few weeks, a new model comes out, and every time the same questions arise: should I switch? what changed what to pay for?
MIT researchers show GPT-6 Astra can improve other AI models by editing their weights
In an experiment, Astra raised Qwen3.5-0.8B's accuracy on MATH-500 from 7.9% to 48.1% by changing only 0.016% of the model's weights.
GPT-6 Astra reportedly edits another AI model's weights without retraining
GPT-6 Astra is claimed to modify only 0.016% of Qwen3.5-0.8B's weights without gradients or retraining, improving reasoning, math results, and instruction following.