Anonymous 'Ox Alpha' Reasoning Model Appears on OpenRouter with Free 1M-Token Context
A stealth model called Ox Alpha has appeared on OpenRouter, offering a 1 million token context window at no cost during its preview period. The model's developer remains anonymous, and OpenRouter says it is acting only as a router, not the model's owner or provider.
Stealth Model Ox Alpha Launches Free on OpenRouter
An unidentified AI developer has released a model called Ox Alpha through OpenRouter, positioning it as a reasoning system built for coding, sustained agentic work, and production workloads. The model is currently free to use and offers a 1 million token context window, according to its OpenRouter listing.
Ox Alpha appeared on the platform on August 20, 2026, though details about its architecture, parameter count, and training data remain undisclosed. OpenRouter explicitly states it is not the model's developer, owner, or provider — it is only routing requests to a third-party operator who has chosen to remain anonymous during this preview phase.
What's Known
According to OpenRouter's listing, Ox Alpha is designed for:
- Long-horizon software engineering tasks
- Complex reasoning workflows
- Multimodal workflows that combine text with visual context
The model supports a 1,000,000 token context window and is currently offered at no cost — no input or output pricing has been set, consistent with a preview or evaluation release. Latency, throughput, and uptime metrics are not yet available due to limited usage data, though OpenRouter reports 100% availability over the last 24 hours across all locations, with three days of uptime history so far.
No benchmark scores, parameter counts, or training cutoff dates have been disclosed. The provider has not confirmed its identity, and no company in the AI industry has yet claimed Ox Alpha as its own.
Data Handling
OpenRouter notes that prompts and completions sent to Ox Alpha are retained by the anonymous provider but are not used for training. All other usage is governed by OpenRouter's Stealth Model Terms, a framework the platform applies to unidentified models during preview windows before a public launch.
What This Means
Stealth releases on OpenRouter have become a common way for AI labs — established or new — to gather real-world usage data and community feedback before a model's public unveiling and commercial pricing. Free access paired with a large context window is a strong incentive for developers to test the model on agentic coding tasks, generating exactly the kind of feedback a lab needs before deciding on final pricing and positioning.
Until the developer is identified, treat any capability claims about Ox Alpha as unverified. The "reasoning model" label and stated focus on coding and agentic workloads mirror positioning used by several major labs for their latest generation of models, but without benchmark data or architectural details, there's no way to independently verify performance claims. Enterprises evaluating the model for production workloads should note that data retention policies, while described as training-free, come from an unnamed party bound only by OpenRouter's Stealth Model Terms rather than a named company's standard enterprise agreement.
Related Articles
Google releases Gemini 4 Neon at $3/$15 per 1M tokens with a 1M-token context window
Google released Gemini 4 Neon, available today through the API at $3 per million input tokens and $15 per million output tokens. It is the fourth new Gemini 4 model in four weeks, with a context window of about one million tokens.
OpenAI releases GPT-Live-1.1 in the API, with standby mode billing idle WebRTC time at $0.02 per minute
OpenAI has released GPT-Live-1.1 in the API. The full-duplex voice model improves multilingual accents and transcription, and a new standby mode bills idle WebRTC time at $0.02 per minute instead of $0.05.
Baseten releases open-weight bison-1 and groundhog-1 decision models; bison-1 scores 76 on AA Decision Index v2.1
Baseten released bison-1 and groundhog-1, two open-weight decision models built on Gemma 4. Bison-1 scores 76 on the AA Decision Index v2.1, and both models support 128k context and image input.
Thinking Machines Lab releases Inkling 2 Mini: 26B-parameter looped MoE with 3B active and 1M-token context
Thinking Machines Lab has released Inkling 2 Mini, a multimodal Mixture-of-Experts model with about 26B physical total parameters and 3B active per token. It repeats its middle layers three times, supports a 1M-token context, and ships with a DSpark drafter. Hugging Face reports it leads 32 of 48 benchmarks against four comparable models.
Comments
Loading...