OpenRouter’s Union Alpha Goes Paid After Demand Overwhelms Capacity

The anonymous coding model’s sudden pricing shift turns a free stress test into a paid service, while its identity and a Cloudflare route give developers more context for deciding whether to use it.

By 2 min read
OpenRouter’s Union Alpha Goes Paid After Demand Overwhelms Capacity
OpenRouter’s Union Alpha Goes Paid After Demand Overwhelms Capacity

Listen to this story

The audio brief

About 1:30
0:001:30
Read transcript
Union Alpha has gone from free to paid after roughly one day, because demand overwhelmed the model’s available capacity. The planned seven-day preview was supposed to cost nothing, but usage reached billions of tokens per minute. AWS reportedly tripled processing capacity overnight and still could not keep up. That turns a curious launch into a useful signal about demand for models built around long coding and agent workflows. Union Alpha accepts text and images, with a context window of 262,144 tokens—roughly the amount of material it can consider in one request—and it can produce up to 131,072 completion tokens. OpenRouter describes it as multimodal and suited to research, coding, and agentic work. The paid API now lists $2.50 per million input tokens, twenty-five cents for cached input, and $7.50 per million output tokens. For agents that repeatedly write, test, and revise code, those separate rates can materially shape the economics. OpenRouter says prompts and completions are not stored or used to train future models. The model remains available through OpenRouter and Cloudflare. Its provider was initially anonymous, although Gigazine reported an identification as Pareto 26.9, from unbiased.ai; OpenRouter still presents the provider as anonymous. The key constraint now is straightforward: whether the model’s long-context capabilities justify paying for access after the free launch exposed more demand than supply.

Story brief

3 key points

OpenRouter has ended Union Alpha’s free preview and switched it to paid access after demand reached billions of tokens per minute and overwhelmed available capacity. The model targets long coding and agent workflows with a 262,144-token context window, multimodal inputs, and up to 131,072 output tokens. Listed pricing is $2.50 per million input tokens, $0.25 for cached input, and $7.50 for output. It remains...

  1. 01

    The planned seven-day, $0 preview lasted roughly one day; AWS reportedly tripled processing capacity overnight but demand still exceeded supply.

  2. 02

    Union Alpha supports text and images, a 262,144-token context window, and up to 131,072 completion tokens.

  3. 03

    OpenRouter lists $2.50 per million input tokens, $0.25 cached input, and $7.50 output.

Union Alpha, the anonymous AI model released through OpenRouter on September 16, has moved to paid access after its planned weeklong free preview ended in a day. Demand rose to billions of tokens per minute, and capacity remained insufficient even after AWS reportedly tripled processing capacity overnight.

That rapid reversal is the central update to a launch built around anonymous access. Union Alpha initially appeared as a “stealth model,” with OpenRouter routing requests to an unnamed third-party provider rather than presenting itself as the model’s developer or owner. It was described as a collaboration involving OpenCode and OpenRouter.

A model built for long coding jobs

The appeal is straightforward for developers working with large amounts of material. Union Alpha accepts text and images, and has a 262,144-token context window—the amount of input it can handle in one request. OpenRouter describes it as a multimodal model for research, coding and agentic workflows, while other launch coverage says it is optimized for coding agents.

The terms developers can act on

  • OpenRouter’s listing says prompts and completions are not stored or used to train future models.
  • The model supports up to 131,072 completion tokens, alongside its 262,144-token context window.
  • Union Alpha is available through OpenRouter and Cloudflare, offering two documented access routes.
The new listed API prices
$2.50Input tokens

The post-preview input price is listed at $2.50 per million tokens.

$0.25Cached tokens

Cached-token pricing is listed at $0.25 per million tokens.

$7.50Output tokens

The post-preview output price is listed at $7.50 per million tokens.

Paid access changes the calculation

The new rates matter most for the workflows Union Alpha is targeting. Agent systems can make repeated model calls as they write, test and revise code, so the separate input, cached-input and output prices make usage costs part of the choice rather than an afterthought. The initial preview had been listed at $0; paid access replaces that temporary zero-price offer with metered API use.

The stealth label no longer hides the same thing

The provider identity has also become less opaque. Gigazine reported that Union Alpha was later identified as Pareto 26.9, a model from unbiased.ai. That identification adds a name to a release that originally asked users to judge the system without knowing its underlying provider; OpenRouter’s own listing still characterizes its role as routing requests to an anonymous third party.

Cloudflare availability gives the model a second documented distribution channel, separate from OpenRouter’s routing layer. For prospective users, the practical questions have narrowed: whether Union Alpha’s long-context, multimodal setup fits their work, and whether the listed paid rates are acceptable after a free launch exposed more demand than the service could sustain.

Sources

  1. openrouter.aiUnion Alpha - API Pricing & Providers
  2. gigazine.netA mysterious high-performance AI model,
  3. cryptobriefing.comOpenCode launches Union Alpha model for free use on OpenRouter

Loading discussion...