How to Use Ox Alpha for Free: 3 Ways

Quick answer: Yes — Ox Alpha is genuinely free right now, and there are three ways to use it: the glm5.app browser chat (zero setup, no API key), the OpenRouter API at $0 per 1M prompt tokens / $0 per 1M completion tokens during the preview period, and OpenCode Go, where Ox Alpha ("Ox Alpha Free") is free for one week with near-unlimited usage. If you just want the least friction, skip the keys and the terminal: open the glm5.app chat page and start typing.

If you've been hunting for "ox alpha free" access, you've probably hit the same wall: one post points at OpenRouter, another at OpenCode Go, a third at some web chat — each with different setup steps, different free windows, and almost nobody explaining what the "free" actually means for your data. This guide sorts the three paths out, gives you copy-paste steps for each, and covers the one thing most tutorials skip: what an anonymous stealth provider means before you paste sensitive code into it.

Everything below is based on the official OpenRouter model listing, the public OpenRouter API, and OpenCode's announcement, all checked on 2026-08-22. We have not run our own benchmarks — Ox Alpha's own model page currently shows no benchmark scores — so treat performance claims, including community ones, with that caveat in mind.

The Short Version: 3 Paths at a Glance

All three routes serve the same model: Ox Alpha (stealth/ox-alpha), released on OpenRouter on 2026-08-20, with a 1,048,576-token (1M) context window, up to 131,072 output tokens, and text + image + video input. Reasoning is always on (mandatory), defaulting to max effort. Choose the path by your workflow, not by the model — it's the same one everywhere.

Path 1: glm5.app Web Chat — The No-Key Route

This is the fastest way to use Ox Alpha free, full stop.

1. Open https://glm5.app/chat?model=stealth/ox-alpha in any modern browser.
2. Type your question or paste your task. That's it — no API key, no billing setup.

What you get out of the box: the full 1M-token context window, the model's reasoning chain (it's always enabled for this model), and multimodal input — you can attach images or video alongside text and get text answers back. That covers the model's intended strengths: complex reasoning, long-horizon tasks, and workflows that mix text with visual context.

If you just want to see the model before committing to any setup, start a free Ox Alpha chat on glm5.app — it's the same model you'd get through the API, minus the ceremony.

Path 2: OpenRouter API — $0/$0 During Preview

For developers and anyone building Ox Alpha into a tool, script, or agent, the OpenRouter API is the direct route — and during the preview it costs literally nothing: $0 per 1M prompt tokens and $0 per 1M completion tokens (the model page simply says "PRICE: Free").

Step 1 — Get a key. Create a free account at openrouter.ai, open the Keys page, and generate an API key. OpenRouter is the gateway here, not the developer: Ox Alpha is built and operated by a third-party provider that has chosen to remain anonymous during this preview, and OpenRouter only routes requests to it.

Step 2 — Call the API. The model ID is stealth/ox-alpha. A minimal request looks like this:

curl https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "stealth/ox-alpha",
    "messages": [
      {"role": "user", "content": "Refactor this function to handle streaming input"}
    ]
  }'

Step 3 — Use the parameters that matter. Per the public API, Ox Alpha supports tool calling (tools, tool_choice) and structured output (response_format), plus temperature, top_p, top_k, max_tokens, include_reasoning, and reasoning_effort (max, high, or low — defaults to max). Defaults are temperature 1 and top_p 0.95.

What to expect operationally (OpenRouter's monitoring, ~3-day window after release): throughput of 24 tokens/s (P50), latency of 5.81s (P50), uptime 99.99% (3d), availability 99.14% (3d). It's a reasoning model, so responses arrive slower than a fast chat model — that's the trade-off for the long CoT-style output.

The heavy users so far are telling: within the first two days, roughly 657B prompt tokens and 7.95B completion tokens were consumed, with the top five apps all agentic coding tools — Hermes Agent (120B tokens), Claude Code (108B), Oh-My-Pi (93.5B), DeepSeek Harness multimodal-bridge (84.8B), and ZCode (51.5B).

Path 3: OpenCode Go — The One-Week Free Window

The third route targets terminal users. OpenCode's announcement (checked 2026-08-21) puts Ox Alpha Free on the OpenCode Go model list, free for the next week, with three notable conditions:

• Near-unlimited usage — the announcement cites a capacity of 100T tokens/day.
• It won't count against your Go usage — your normal Go allowance is untouched while the promo runs.
• Zero data retention — a meaningful difference from the other two paths, more on that below.

You also get the same 1M context and multimodal input, which fits Ox Alpha's stated positioning as a stealth model for coding and agentic work. If your daily driver is a terminal-based coding agent, this is the most natural fit — and the zero-retention guarantee is the strongest privacy posture of the three options, at least while the window lasts.

How Long Is "Free" Anyway? Two Different Clocks

The biggest source of confusion is that the two free offers run on different timers:

• OpenRouter preview pricing — the $0/$0 rate has been live since the model's release on 2026-08-20 and has no announced end date. "Preview" means the provider can change it at any time; post-preview pricing is not disclosed.
• OpenCode Go free week — announced 2026-08-21 as "free for the next week," which puts the window ending around late August. After that, usage reverts to normal Go accounting.

glm5.app's web chat rides on the same preview availability. None of the three paths is a permanent free tier — treat them as a trial window and check the official pages before building a production dependency on them.

Free ≠ No Cost: What the Anonymous Provider Means for Your Data

This is the part most "how to use Ox Alpha free" posts skip, and it matters more than the setup steps. The official listing states plainly:

• Ox Alpha is a stealth model, developed and operated by a third-party provider who has chosen to remain anonymous during this preview.
• OpenRouter routes requests to it and is not its developer, owner, or provider.
• Prompts and completions are retained by the provider and are not used for training; all other use is governed by the Stealth Model Terms.

Read that again: on the OpenRouter path, free pricing does not come with a no-retention promise. Practical implications:

1. Don't paste secrets. No proprietary source code, customer data, credentials, or anything you couldn't accept being retained by an anonymous third party. The provider's identity, architecture, and parameter count are all not disclosed — you can't audit who's holding what.
2. Zero retention is the exception, not the rule. It's been announced for the OpenCode Go path specifically, and even that is a one-week promotion. On the OpenRouter route, "not used for training" is the only commitment — retention is explicitly allowed.
3. There are no official benchmarks. The model page shows no intelligence/coding/agentic scores, and Artificial Analysis had not listed Ox Alpha as of 2026-08-22. Community-reported numbers (e.g., roughly 80% on a 10-task DeepSWE subset, or a Kingbench result of 87.5%) are community-reported and not independently verified — the samples are tiny and the benchmarks are non-standard. Test it on your own workload before you trust anyone's score.
4. The terms can change. Anonymous provider + preview period + undisclosed pricing means the deal on offer today may not be the deal next month. For regulated or confidential work, wait for a disclosed provider or use a model whose retention terms you can actually review — the Stealth Model Terms page is the only binding document, and it's worth a read before you wire this into anything serious.

None of this makes Ox Alpha a bad deal — a 1M-context reasoning model at $0 is remarkable while it lasts. But "free" here means free of charge, not free of conditions.

FAQ

Is Ox Alpha really free? Yes. OpenRouter lists it at $0 per 1M prompt tokens and $0 per 1M completion tokens during the preview ("PRICE: Free"), OpenCode Go offers Ox Alpha Free for one week with near-unlimited usage, and glm5.app provides a free browser chat. Post-preview pricing is not disclosed.

How long will Ox Alpha stay free? The OpenCode Go free week runs from the 2026-08-21 announcement for roughly one week. The OpenRouter preview pricing has no announced end date — it's "preview" pricing, so the provider can change it at any time.

Do I need an API key to use Ox Alpha? Only for the OpenRouter API path — you need a free OpenRouter account and key. The glm5.app browser chat needs no key at all, and OpenCode Go is accessed through the Go service.

Does Ox Alpha train on my data? Officially, no: per the OpenRouter listing, prompts and completions are retained by the provider and are not used for training. But retention is allowed, and all other use is governed by the Stealth Model Terms — so treat anything you send as potentially retained and don't paste sensitive material. The OpenCode Go path announced zero data retention during its free window.

What is the easiest way to use Ox Alpha? The glm5.app browser chat — no key, no account, no code. Open the chat page and type.

Is Ox Alpha good for coding agents? That's exactly what it's positioned for — a reasoning model "designed for coding, sustained agentic work, and production workloads" — and the early usage data backs it up: the top five apps by token volume are all agentic coding tools. It supports tool calling, structured output, and a 1M context window, and the OpenCode Go route is built around terminal use.

If you want the fastest possible look at the stealth model, use Ox Alpha free on glm5.app — one link, no key, and the full 1M-context reasoning model in your browser. For the API route, grab the stealth/ox-alpha model ID and the curl example above and you're live in minutes. Either way, mind the retention terms, treat the free windows as temporary, and check the official pages before you rely on it in production.

Last updated: August 22, 2026

Sources

• OpenRouter — Ox Alpha model page
• OpenRouter Models API
• Stealth Model Terms
• OpenRouter Docs
• OpenCode Docs — Models