Antimatter

Pricing

Every model, priced in the open

You pay the provider's published price plus 10%. That is the whole margin, on every model.

281 to 286 of 286 models (of 336)

ModelPrice per 1MInput per 1MOutput per 1MLast 90 daysContextMarginSpeed
Lyria 3 Clip Previewgoogle/lyria-3-clip-previewvisionjson, declaredmakes audiofree infree outfreefreeno history yet1Mnoneunmeasured
Ling 3.2 Flash (free)inclusionai/ling-3.2-flash:freevisiontools, declaredjson, declaredfree infree outfreefreeno history yet1Mnoneunmeasured
ERNIE-5.5-Flash-Preview (free)baidu/ernie-5.5-flash-preview:freetools, declaredjson, declaredfree infree outfreefreeno history yet1Mnoneunmeasured
Flint (free)recursive/flint-v1.0:freetools, declaredjson, declaredfree infree outfreefreeno history yet262knoneunmeasured
Dots3-Note Preview (free)dots-studio/dots-3-note-preview:freevisiontools, declaredjson, declaredfree infree outfreefreeno history yet512knoneunmeasured
Jev Routertypesafe/jev-routervisionreads audioreads filesreads videotools, declaredjson, declaredn/an/an/ano history yet1Mn/aown key onlyunmeasured

“list” is the provider's published rate, read from its catalog. A tag with a tick was measured working by a probe here. One without is the provider's own claim. First token is the median measured here over the last week, from real traffic and probes. Prices follow the provider. When it changes a rate, this list and your bill change with it.

Calculator

Cost of a request

The three cheapest models in the list as filtered now. Add up to four of your own with Compare. This uses the same function the gateway bills with. You set the token counts, and the tasks are round numbers to start from.

how many
What a thousand of 1,200 tokens in and 350 tokens out costs on each model
ModelAntimatterMargin
Ling 3.0 Tinyinclusionai/ling-3.0-tiny$0.0371$0.00338
Llama 3.1 8B Instructmeta-llama/llama-3.1-8b-instruct$0.0418$0.0038
Mistral Nemomistralai/mistral-nemo$0.0498$0.00453

A thousand requests of 1,200 tokens in and 350 out. A cached prompt token costs less where the provider publishes a cache rate.

The margin
333 priced models, by what a token costs (three in to one out) and how much each can hold. 3 more with no published price are in the list but not on the axis. Out of 336 models. The list below has all of them, with search.

margin over the provider's price

+10%

You pay the provider's published price plus 10%. That is the whole margin, on every model.

It is the same for every account. There is no fee to add credit, so what you send is what you are credited. This is not cheaper than buying direct. You pay the provider's price plus this margin for one key, one API and every model.

models
336
priced
333
labs
60

Autopilot

Let Autopilot choose

An Autopilot id picks a model for each request. It is billed at that model's price from this list, with the same margin. The answer names the model in x-antimatter-model.

  • antimatter/autoBalances price, measured speed and measured quality.
  • antimatter/auto:cheapThe lowest blended price in the catalog.
  • antimatter/auto:fastThe lowest measured time to first token.
  • antimatter/auto:codeModels whose code passed the gateway's own tests, with enough context for code.
  • antimatter/auto:extractModels measured pulling fields out of text exactly.
  • antimatter/auto:reasonModels measured solving short puzzles correctly.
  • antimatter/auto:longA long context window, from the catalog.
  • antimatter/auto:visionA model that takes images as input.
  • antimatter/auto:toolsA model whose tool calls were measured working.
  • antimatter/auto:jsonA model whose structured output was measured working.
Keys are free. A lifetime allowance of $0.25 covers the first requests of an account that has never deposited.
We take only $ANTI right now: ANTI on Robinhood Chain. Send it, then claim the transaction from your account. It is credited in dollars at the credit price when you claim it.
Agents use a key made with a wallet signature. Paying per request with x402 is not on here. See how
Data retention explains what the API keeps, for how long and how to remove it.

Start with a key

A key is free to create on the dashboard, and a new account's first requests are covered by a free allowance of $0.25. After that, requests are billed at the prices above.

Coming from OpenRouter? What changes and what stays the same.