Models and prices

What we serve, how much context each model takes, and what it costs per million tokens.

cog-fast256K context

High-volume model for agents, pipelines and long-document extraction where a refusal mid-run is a failed run: security automation, OSINT, moderation datasets, batch translation of hostile source material.

reasoning
tool-use
file-input
json-mode

cog-pro256K context

Dense model for one hard task at a time: exploit and malware analysis, long-form fiction with dark themes, investigative summaries, clinical and harm-reduction drafting that needs the complete answer. Served without a moderation layer.

reasoning
tool-use
file-input
json-mode

cog-embed32K context

Retrieval embeddings, 1536 dimensions, for RAG over your own corpora, including material a hosted embedding API would refuse to index. Text in, vector out, priced per input token.

embedding

Prices

Rates are per million tokens, in US dollars, and are identical on every plan. Prices are billed as credits in the console, where one credit is one cent.

Model	Context	Max output	Input /M	Cached input /M	Output /M
`cog-fast`	256K	32K	$0.600	$0.060	$3.00
`cog-pro`	256K	32K	$1.20	$0.120	$5.00
`cog-embed`	32K	n/a	$0.020	n/a	n/a

Cache reads are billed at a tenth of the input rate. Writing to the cache costs 1.25 times the input rate, once, and then every later read of that prefix is billed at the cached rate. For an agent that resends the same long system prompt on every turn, that is the difference between paying for the prompt once and paying for it every time.

Context and output ceilings

Model	Context window	Maximum output
`cog-fast`	262,144	32,768
`cog-pro`	262,144	32,768
`cog-embed`	32,768 input	not applicable

The context window is input plus output. A request over the ceiling is rejected rather than truncated, so a prompt is never silently cut in half.

Input ceiling on cog-pro by plan

cog-fast serves its full window on every plan. cog-pro reads further as the plan goes up:

Plan	Maximum input on `cog-pro`
Pay as you go	64K
Builder	128K
Pro	256K
Max	256K
Reserved	256K
On-prem	256K
Any plan, `cog-fast`	256K

Choosing between them

cog-fast is the default for anything that runs unattended: agents, pipelines, extraction over many documents, moderation datasets, batch translation. It is the model to use when a stall costs more than a worse answer.
cog-pro is for a single hard task where the whole answer has to be present: exploit analysis, long-form writing, investigative summaries, clinical or harm reduction drafting.
cog-embed turns text into 1,536-dimension vectors for retrieval over your own corpus.

List models

curl https://api.cognitivers.com/v1/models \
  -H "Authorization: Bearer $COGNITIVERS_API_KEY"

Returns the ids your key is scoped to. Clients that support discovery read this endpoint instead of requiring a hardcoded list; on those clients you only supply the base URL.

Availability

A model is listed here only once it is served. Nothing on this page is a preview, a waitlist or a roadmap item, so if an id appears above, a request with it will be answered.