model release

Anonymous Provider Launches Union Alpha, a Free 262K-Context Multimodal Model on OpenRouter

TL;DR

A third-party provider using the alias 'Stealth' has released Union Alpha on OpenRouter, a multimodal model with a 262K context window, currently free to use during its preview period. The model's developer remains anonymous, and OpenRouter states it is not the model's owner or operator.

2 min read
0

Union Alpha Debuts as Free Preview on OpenRouter

A new multimodal model called Union Alpha appeared on OpenRouter on September 16, 2026, listed under the identifier stealth/union-alpha. The model's actual developer has chosen to remain anonymous during this preview period. OpenRouter, which routes inference requests to the model, has clarified it is not the model's developer, owner, or provider.

Union Alpha ships with a 262,000-token context window and is currently free to use through OpenRouter. According to the model's listing, it targets research, coding, and agentic workflows while aiming for what the provider calls "frontier-level performance across a broad range of general-purpose tasks." No independent benchmark scores, parameter count, or training cutoff date have been disclosed.

Performance Metrics

OpenRouter's own telemetry provides some operational data, though these figures reflect infrastructure performance rather than model quality:

  • P50 latency: 5.39 seconds
  • Throughput: 18 tokens per second
  • Uptime (3-day): 99.85%
  • Availability (3-day): 98.14%

These numbers describe how reliably OpenRouter can route and serve requests to the model, not its reasoning or coding capability. No standard benchmark results (MMLU, HumanEval, or similar) have been published for Union Alpha.

Data Handling

According to OpenRouter, prompts and completions sent to Union Alpha may be retained by the anonymous provider but are not used for training. All other usage is governed by OpenRouter's Stealth Model Terms, a separate policy applied specifically to unidentified model providers on the platform.

Part of a Broader Stealth Pattern

Union Alpha is not an isolated listing. The same anonymous provider (or providers operating under the "Stealth" branding) has also released Ox Alpha, a reasoning-focused model with a 1 million-token context window aimed at coding, long-horizon agentic work, and production use cases involving text and visual context. The pairing suggests a coordinated release strategy: one multimodal generalist model (Union Alpha) and one long-context reasoning specialist (Ox Alpha), both launched without public attribution.

Stealth releases on OpenRouter are a recognized pattern in the industry, often used by labs testing unreleased checkpoints under real-world traffic before an official launch and branding announcement. Free pricing during this phase is typical, functioning as an incentive for developers to generate usage data and feedback ahead of a formal release with paid tiers.

What This Means

Union Alpha's claims of "frontier-level performance" cannot be verified — there are no published benchmark scores, and the operational metrics available (latency, throughput, uptime) say nothing about output quality. The 262K context window is competitive but not exceptional against known models with 1M+ token windows. Until the provider identifies itself and publishes verifiable evaluation results, treat Union Alpha as an unproven entrant worth testing in low-stakes settings, not a model to build production systems around. The free pricing window is likely temporary and will end once the provider reveals its identity or moves to general availability.

Related Articles

model release

ScaleDown launches Summarize, a 1M-context abstractive summarization model at $0.04 per 1M input tokens on OpenRouter

ScaleDown has released Summarize, an abstractive summarization model with a 1M-token context window, listed on OpenRouter on October 13, 2026. OpenRouter shows $0.04 per 1M input tokens (20% off the $0.05 list price) and $0 per 1M output tokens. Parameter count and benchmark scores have not been disclosed.

model release

StepFun releases Step 5 Preview: 600B MoE with 1M context at $1/$2.70 per 1M tokens

StepFun has listed Step 5 Preview, a sparse Mixture-of-Experts model with 600B total and 27B active parameters and a 1.0M-token context window. It is priced at $1 input and $2.70 output per 1M tokens on OpenRouter. StepFun positions it as its flagship model for agentic work.

model release

Unbiased releases Pareto 26.9: 262K-context composite model at $2.50/$7.50 per 1M tokens

Unbiased has released Pareto 26.9, a multimodal composite model with a 262K-token context window, priced at $2.50 per 1M input tokens and $7.50 per 1M output tokens. It is listed on OpenRouter with a Oct 13, 2026 release date. A cheaper 1.0M-context preview, Pareto 26.10, is already listed alongside it.

model release

Microsoft releases Decision-1, a Qwen3.5-9B-based model for classification and routing, at $0.042 per 1M input tokens

Microsoft has released Decision-1, a decision model built on Qwen3.5-9B for classification, evaluation, and routing. Microsoft claims 83.5% accuracy across 36 benchmarks and 85 ms latency. Input tokens cost $0.042 per million, and output tokens are free.

Comments

Loading...