Switching from OpenRouter
Change two lines, keep the rest
Antimatter speaks the same OpenAI compatible wire and takes the same vendor/model ids, so moving an OpenRouter integration is the base URL and the key. Below: what stays the same, what is different, and how to compare prices model by model.
The two lines
Point your client at Antimatter and use an Antimatter key. Nothing else in the request changes.
# was https://openrouter.ai/api/v1
base_url = "https://useantimatter.com/v1"
# was sk-or-...
api_key = "am_live_..."The key comes from the dashboard; it is shown once.
The same request, end to end:
from openai import OpenAI
client = OpenAI(
base_url="https://useantimatter.com/v1",
api_key="am_live_...",
)
reply = client.chat.completions.create(
model="google/gemini-4-argon",
messages=[{"role": "user", "content": "Hello"}],
)
print(reply.choices[0].message.content)What stays the same
- Model ids: the same vendor/model ids, 471 worlds today, listed by
GET /v1/models. - The OpenAI chat completions wire at
/v1/chat/completions, so the official openai clients work with a new base URL. The Anthropic messages wire and OpenAI's Responses wire are served too. - Streaming as server sent events, in the same chunk format, with usage on the last chunk.
- Tool calling and structured output (
response_format), translated where a vendor speaks another wire.
What is different
- Price
- Each request is billed at the upstream's published price plus 10%, the same on every world and for every account, stated on every row of the price list.
- Signing in
- With a wallet signature, or with email, Google or X, which makes a wallet for you. There is no password.
- Paying
- Credit is added in $ANTI right now, not stablecoins: ANTI on Robinhood Chain. It is credited in dollars at the credit price when it is claimed.
- Free to start
- A new account's first requests are covered by a lifetime free allowance of $0.25.
What Antimatter adds
- Autopilot:
antimatter/autoand 9 task modes choose a world per request from price, measured speed and measured capability. - Comparable fallback:
x-antimatter-fallback: comparablehands a failed request to the nearest model that costs no more. - Strict JSON:
x-antimatter-json: strictchecks the answer against your schema and asks once more when it does not hold. - Resumable streams: a stream opened with
x-antimatter-resumable: 1runs to its end if the connection drops, and can be read again. - The batch API:
/v1/filesand/v1/batches, as OpenAI's clients send them.
Comparing prices, model by model
There is no blanket saving to quote, and this page claims none. Every row of the price list shows two rates for the same id: the one the upstream publishes and the one billed here. To compare with what you pay now, put the rate on your current bill beside the billed rate for the same id.
| World | Upstream, in / out | Billed here, in / out | Margin |
|---|---|---|---|
| Gemini 4 Argongoogle/gemini-4-argon | $2.00 / $10.00 | $2.20 / $11.00 | +10% |
| Claude Haiku 5.5anthropic/claude-haiku-5.5 | $0.10 / $0.50 | $0.11 / $0.55 | +10% |
| Mistral Large 4mistralai/mistral-large-4-0 | $0.68 / $2.09 | $0.748 / $2.299 | +10% |
| GPT-6.1 Sol Proopenai/gpt-6.1-sol-pro | $2.00 / $10.00 | $2.20 / $11.00 | +10% |
| GLM 5.3 Primez-ai/glm-5.3-prime | $2.80 / $8.80 | $3.08 / $9.68 | +10% |
| Qwen3.8 Max Primeqwen/qwen3.8-max-prime | $4.00 / $12.00 | $4.40 / $13.20 | +10% |
Every world on the price list, with search and a calculator.