Ox Alpha on OpenCode: Long Context and Free Preview Configuration

One-sentence takeaway

OpenCode presented Ox Alpha as a one-week free stealth preview with 1M context, multimodality, and zero data retention, making it useful for long-task prototypes but not a long-term pricing or SLA commitment.

Use cases

• Suitable tasks: Long-context repositories, mixed text/image/video input, and low-cost agent prototypes.

• Unsuitable tasks: Inferring a personal quota from “100T tokens per day,” or treating a zero-data-retention claim as a complete compliance conclusion without checking the current terms.

• Applicable model version: The Ox Alpha stealth preview in the OpenCode announcement; weights and a formal version number are not public.

• Applicable clients, agents, or APIs: OpenCode; the announcement describes the OpenCode entry point.

• Recommended reasoning levels and parameters: The announcement does not disclose temperature, effort, tool permissions, or rate-limit details.

Ready-to-use content

The official post does not publish API JSON or a system prompt. The following configuration checklist is reusable:

Model: Ox Alpha (OpenCode stealth entry point)
Working mode: prioritize long-context, sustained-agent, and multimodal prototypes
Validation: record client version, model identifier, context length, retries, actual latency, and rate-limit events
Data boundary: recheck the current OpenCode/provider zero-data-retention terms before submitting data; do not upload production secrets

Original information and boundaries

• The announcement says: free for the next week, 1M context, multimodal, zero data retention, generous rate limits, and capacity for 100T tokens per day.

• These are OpenCode release claims, not an independent capability benchmark; the free period, capacity, availability, and terms can change.

• OpenRouter separately lists an anonymous Stealth provider; the OpenCode entry point does not confirm the developer's identity.