One key for every model, at wholesale per-token prices
Why one base URL and one balance beats a console per developer: the models on the key today, how the price compares with each developer’s list, and what switching costs in code (one string).
Every model developer runs its own console, its own key format, its own billing and its own idea of a minimum. A team that wants Claude for one job, GPT for another and an open-weight model for the bulk work ends up with three accounts and three invoices. RelaySwitch is one base URL and one prepaid balance in front of all of them, with every model routed through our wholesale network and priced per token. This post is the inventory, read live, and the one-string change in code.
- Chat models
- 469
- Developers
- 47
- Media models
- 101
- Max under list
- 90%
01What is on the key today
469 chat models from 47 developers, the largest being OpenAI (84), Qwen (57), Google (community tune) (34), Google (30), xAI (29). 136 of them are uncensored or abliterated builds, on the uncensored page. 101 image, video, speech and 3D models are on the same key, priced per unit, on the media page.
Every figure on this site is read from the catalog when the page is built, so the counts above are the counts now, not the counts at launch.
02Wholesale, against each list price
76 of the chat models carry the developer's own published list price in the catalog, so the saving can be shown as a percentage rather than claimed. The flagships:
| Model | Ours in / out | List in / out | Under list, out |
|---|---|---|---|
| Claude Opus 5.5 | $2.80 / $10.00 | $4.00 / $20.00 | 50% |
| GPT-5.5 | $1.80 / $9.00 | $5.00 / $30.00 | 70% |
| DeepSeek V4 Pro 0423 | $0.792 / $2.376 | $1.32 / $3.96 | 40% |
| Grok 4.7 | $0.60 / $1.80 | $2.00 / $6.00 | 70% |
| GLM 5.3 | $0.3276 / $1.0296 | $1.092 / $3.432 | 70% |
| Kimi K3 | $0.90 / $4.50 | $3.00 / $15.00 | 70% |
| Qwen3.7 Max | $0.6471 / $1.9412 | $1.475 / $4.425 | 56% |
| MiniMax M3 | $0.09 / $0.36 | $0.30 / $1.20 | 70% |
USD per million tokens. The price index lists every model with its list price, downloadable as JSON and CSV.
Where no list price is on record (most open-weight builds) the table on the price index shows a dash rather than a guess.
03One balance, every method
The balance is USD credit, topped up by Card, Apple Pay, PayPal, USDT, USDC, BTC, ETH, SOL. It moves only when you call, at the per-token price on the model page, input and output separately, cached input cheaper where the model supports it. No subscription, no seat, no minimum spend; credit does not expire.
04The change in code
# three consoles, three keys, three billsclient_a = OpenAI(api_key=os.environ["OPENAI_API_KEY"])client_b = Anthropic(api_key=os.environ["ANTHROPIC_API_KEY"])client_c = genai.Client(api_key=os.environ["GEMINI_API_KEY"])# one base URL, one key, one balanceclient = OpenAI(base_url="https://api.relayswitch.io/v1", api_key=os.environ["RELAYSWITCH_KEY"])for model in ["claude-sonnet-5-5", "gpt-5.5", "gemini-4-argon", "deepseek-v4-pro"]: reply = client.chat.completions.create(model=model, messages=[{"role": "user", "content": "Hi"}])The model id is the developer's id, the request and response keep the OpenAI shape, streaming and tool calls included. The Claude ids also answer the Anthropic Messages format, so the Anthropic SDK and Claude Code work with the same two values; the SDK guide and the Claude Code guide show each.
Every model, one key, one balance
Prepaid from $10. Card or crypto, no KYC.