teamToken vs OpenRouter: the alternative that also generates video and images
OpenRouter is the default choice for LLM routing and hard to beat on catalog size. It does not generate video, and it publishes no per-second pricing because there is nothing to price that way. That gap is what this page is about, along with an honest section on where OpenRouter beats us.
One integration instead of two
If your stack needs an LLM gateway and a media API, that is normally two vendors, two keys and two invoices. Here it is one of each: 10 text models and 51 media models answer on the same key, video billed by the second from $0.0067 and images from $0.02 apiece.
Every one of those prices sits on a public page with a calculator and a comparison against dedicated media APIs. No login, no sales call, no "contact us for pricing".
The text side, without decoration
| What you called | What to use here |
|---|---|
| GPT class, general work | gpt-5.5, gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna |
| Cheaper GPT, bulk work | gpt-5.4, gpt-5.4-mini |
| Coding agents | gpt-5.3-codex, codex-auto-review |
| Gemini | gemini-3.1-pro, gemini-3.5-flash |
| Claude | not available right now; the upstream fell through |
| Open-source and niche models | not offered, and not planned |
The list is pulled from the live catalog at build time rather than typed by hand.
Two rows there say no, and we are leaving them in. Against three hundred models on the other side, pretending otherwise would not survive the first click.
Switching takes one line
For any OpenAI-compatible SDK, the address and the key change and the calls do not.
# before
base_url = "https://openrouter.ai/api/v1"
# after
base_url = "https://api.teamtoken.store/v1"For Codex CLI it is a config block:
# ~/.codex/config.toml
[model_providers.teamtoken]
name = "teamToken"
base_url = "https://api.teamtoken.store/v1"
env_key = "TEAMTOKEN_API_KEY"
wire_api = "chat"What it costs to move money in
Card, SBP or crypto, from $1 by card and $5 in crypto, with the payment provider's percentage printed next to it before you confirm. OpenRouter charges 5.5% on card top-ups and 5% on crypto while passing inference through at 0%. Same business model, moved from the tokens to the till. We do not publish our markup as a separate figure and will not pretend there is none.
Questions we get
base_url and the key change.When OpenRouter is the better call
On catalog breadth it is not close. Three hundred models against our 10 text ones, with open-source weights, niche fine-tunes and a choice of providers per model. If what you need is one specific open model, that decides it.
The 0% inference markup is the most transparent structure on the market if you would rather pay a top-up fee than an invisible margin. They have years of track record, enterprise limits up to $25,000 a month on pay-as-you-go and $200,000 on enterprise, and BYOK for teams bringing their own provider keys.
And the thing worth knowing in both directions: since autumn 2025 Hacker News has carried a running discussion that some providers in their routing serve quantized models under full-size names. OpenRouter acknowledged the concern and shipped an opt-in :exacto flag. The community that caught it is one we do not have yet.