PRO-1 API

One API.
A broader intelligence.

Use one OpenAI-compatible model endpoint for complex reasoning, research, documents, multimodal workflows and production applications.

Start building View documentation

PRO-1 checks each answer for completeness against an explicit rubric before returning it. The methodology and raw per-question data are open.

Live benchmarksprogress published openly
OpenAI-compatiblemodel endpoint
Estimatebefore execution

In Beta — features and pricing may change.

api.promanos.com Ready

Request

POST /v1/chat/completions
Authorization: Bearer $PROMANOS_API_KEY

{
  "model": "pro-1",
  "messages": [
    {"role": "user", "content": "Assess the launch strategy for a B2B AI API and find the risks that are easy to miss."}
  ]
}

01UnderstandTask, files and constraints mapped

02ResearchEvidence and relevant context processed

03VerifyCoverage checked before final answer

Response

The launch strategy is viable, but the highest-risk assumptions are proof, onboarding friction and price comprehension.

PRO-1 · Chat Example

What problem does PRO-1 solve?

Reasoning

The AI market is vast and fragmented.
Today
One task needs a reasoning model, another a vision model, another speech, audio or realtime, another a fast, low-cost layer. New models appear constantly, pricing and quality shift, and developers have to choose providers, integrate different APIs, test outputs, manage cost, and keep rebuilding their AI stack.
For complex tasks, one model is often not enough. You need to break the task down, try several approaches, work across modalities, check completeness, and synthesize the final result.
With PRO-1
PRO-1 is not a selection layer. It is an amplification layer.
To you, PRO-1 is one OpenAI-compatible API. Inside, it uses the strongest capabilities of leading AI systems as components: it understands the task, activates the capabilities it needs — reasoning, vision, speech, audio, realtime, generation and analysis — combines their contributions, checks completeness, and assembles one final result.
So on complex tasks PRO-1 can produce a deeper, more complete answer than a single model alone — that is exactly what the built-in completeness check is for. The methodology and per-question data are open.
No single model wins every tested task — leading systems are complementary. In historical evaluations PRO-1 covered more of the rubric than the tested single-model baselines, but those source runs predate build-stamp capture and do not verify the current production build.
Reliability also varies: in our test on 350 hard prompts, one leading model refused 12% of requests, another 0%. PRO-1 reduces dependence on any one system's failure modes.

modelpro-1processed tokens2,140price$0.05

PDFcontract.pdf

Summarize the key risks in this contract.

Three clauses need attention: auto-renewal, the liability cap, and the data-retention terms.

modelpro-1files1price$0.06

Generate a few product hero options — mint green, studio light.

/v1/imagespro-1images3 · 1024²price$0.12

Make a 5-second looping product clip.

5s · 1080p

/v1/videospro-1length5sres1080p

Hey — can you walk me through today's numbers?

Live · 00:03 Sure — revenue is up 12% week-over-week, driven by activation…

Speech synthesis

“Hi — this is PRO-1's synthesized voice.”

/v1/realtime · /v1/audiopro-1latency320 msaudiolive + TTS

Help me plan our launch.

One quick question first — what are we launching?

modelpro-1interactiveclarify

Model endpoint

One model endpoint

Use pro-1 through the existing API surface.

Compatibility

OpenAI-compatible

Change the base URL, API key and model name.

Benchmark proof

Open methodology

Per-question data published; live progress at /v1/benchmarks/live.

Cost control

/v1/estimate

Preview supported request cost before execution.

Benchmarks

How PRO-1 is evaluated.

PRO-1 is evaluated on rubric coverage and no-regret behavior — one strict checker epoch, decoy control, and a separate contamination-resistant exactness track. Every per-question result and the methodology are public.

Methodology

Rubric coverage, strict checking, open data

Each answer is scored on how completely it covers a rubric of required points, under a single strict checker epoch with a decoy control that measures the checker's false-positive rate. A separate contamination-resistant track re-checks exactness on fresh, previously unpublished questions. Every per-question result is published machine-readable.

The historical July 2026 runs did not capture a build stamp, so they are not proof of the current production build — their scores are not shown on this page. Each linked proof endpoint states its own proof_status and era note; numbers will return here only from a run bound to the running build. Methodology was tightened in July 2026 — figures are not comparable with earlier publications.

Open the per-question data Open Fresh No-Regret · contamination-resistant Open the live benchmark progress

Quickstart

Up and running in a minute.

PRO-1 is /v1/chat/completions with model: "pro-1". Works with the official OpenAI SDK — just change the base_url and the key.

Demo: cURL, Python (OpenAI SDK), Node

Keep your key on the server. The model returns the cost and version in the response.

copy

You can preview and copy the code.

curl https://api.promanos.com/v1/chat/completions \
  -H "Authorization: Bearer $PROMANOS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "pro-1",
    "messages": [
      {"role": "user", "content": "Assess the launch strategy for a B2B AI API and find the risks that are easy to miss."}
    ]
  }'

# See the cost/time range BEFORE you run (without calling providers):
curl -X POST https://api.promanos.com/v1/estimate \
  -H "Content-Type: application/json" \
  -d '{"model": "pro-1"}'

# pip install openai
from openai import OpenAI

client = OpenAI(
    base_url="https://api.promanos.com/v1",
    api_key="promanos_...",  # token from /account
)

resp = client.chat.completions.create(
    model="pro-1",
    messages=[
        {"role": "user", "content": "Assess the launch strategy for a B2B AI API and find the risks that are easy to miss."}
    ],
)

print(resp.choices[0].message.content)

// npm install openai
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.promanos.com/v1",
  apiKey: process.env.PROMANOS_API_KEY, // token from /account, keep it on the server
});

const resp = await client.chat.completions.create({
  model: "pro-1",
  messages: [
    { role: "user", content: "Assess the launch strategy for a B2B AI API and find the risks that are easy to miss." },
  ],
});

console.log(resp.choices[0].message.content);

{
  "mcpServers": {
    "promanos-docs": {
      "type": "http",
      "url": "https://api.promanos.com/mcp/docs"
    },
    "promanos": {
      "type": "http",
      "url": "https://api.promanos.com/mcp",
      "headers": { "Authorization": "Bearer sk-your-token" }
    }
  }
}Connect remotely: public read-only documentation, or authenticated PRO-1 and Marketplace tools for IDEs and agents. Docs MCP needs no key. Product MCP uses your PRO-1 API token and the same billing as the API. Require approval for order actions.

Create a key at /account

Keep your key on the server. An empty balance returns a clear 402 — not a silently degraded answer.

Capabilities

A single interface for demanding AI work.

Core model capabilities are separated from the application patterns you can build on top of them.

Core capabilities

Advanced reasoning

Work through complex questions with multiple approaches, synthesis and completeness checks.

Web research

Collect current evidence, compare sources and return an answer grounded in relevant context.

Documents and PDF

Extract, analyze and answer from contracts, reports and other attached documents.

Vision and multimodal input

Understand screenshots, diagrams, images, text, documents and audio in one workflow.

Code execution

Use code and math tools for calculation, transformation and result verification.

Image and video generation

Create visual media through the same API family used by the rest of your application.

Realtime and voice

Build low-latency voice experiences, transcription and speech-enabled applications.

MCP and external tools

Connect trusted tools and external systems when the task needs action beyond the model.

Clarification and verification

Ask for missing constraints and check coverage before returning the final result.

Structured output

Return predictable machine-readable results for product and automation workflows.

Application patterns

Research agents

Evidence-led market, strategy, policy and due-diligence systems.

Document intelligence

Extraction, comparison, review and question answering across document collections.

Complex workflows

Multi-step assistants that combine reasoning, tools, media and verification.

Pricing

Prepaid, pay-as-you-go pricing.

Pay as you go, in the standard $/1M format. We bill processed tokens (input + full output, including reasoning tokens). Billing is by actual usage; a pre-run cost range is returned by /v1/estimate.

No subscription and no promotional credits. Fund your account before the first request; actual usage is deducted from your prepaid balance.

Pricing version: pro-1-usd-2026-07-11-r2 · effective 2026-07-11 · minimum $0.002

Usage rates

Input tokens

$12

per 1M tokens

Everything PRO-1 processes as input: the question and all the context you pass.

Output tokens

$48

per 1M tokens

The full output: generation, synthesis and verification of the answer (reasoning tokens).

Cost examples

$0.36-$3.72

by processed input and output tokens

10k input + 5k output = $0.36
50k input + 15k output = $1.32
150k input + 40k output = $3.72

All capabilities & prices

Service	What it does	≈ adds to the request
Built-in tools — billed as tokens, no separate fee
Web search	up-to-date facts (weather / rates / news) with sources	≈ +$0.01–0.03
Code execution	code and math in a sandbox + result verification	≈ +$0.01–0.03
Files / PDF	text extraction from attachments	$12 / 1M tokens · by size
Vision (images)	image analysis with a vision model	$12 / 1M tokens · by size
Audio transcription	speech from an audio attachment to text (speech-to-text)	$12 / 1M tokens · by size
Video understanding	analyze a video — sampled frames + the audio track	$12 / 1M tokens · by length
Generation & audio — separate endpoints
Image generation	text → image (standard & premium)	from ≈ $0.04 / image
Video generation	text → a short video clip	from ≈ $0.40 / clip
Speech synthesis (TTS)	text → natural speech, OpenAI-compatible	from ≈ $0.015 / 1k chars
Voice cloning	a branded voice from your samples	from ≈ $0.20 / voice
Realtime voice	live, low-latency voice conversations	by audio usage

Generation and audio are cost-plus. /v1/estimate returns a pre-run range; the response reports the final price based on actual usage.

FAQ

Frequently asked questions.

The essentials before you integrate.

Is PRO-1 one model or multiple models?

To you, PRO-1 is one model and one API. Inside, PRO-1 uses its own intelligent layer to combine the strengths of several leading AI systems, verify completeness, and synthesize the final result. The exact internal composition — providers, models, count, and selection logic — is proprietary.

Is it really OpenAI-compatible?

Yes. Point the official OpenAI SDK at https://api.promanos.com/v1, set the model to pro-1, and use your API key — no other code changes.

How is pricing calculated?

Per processed token — $12 / 1M input and $48 / 1M output — billed from a prepaid balance. There are no subscriptions or promotional credits. Fund the account before use; /v1/estimate returns a cost range before you run.

What happens to my data?

Accounts and data are isolated per user and stored in a single region; your content is not used to train models. See the Privacy Policy for details.

Are there rate limits?

There's a per-minute request limit, set by your usage tier (level). The tier rises automatically as your total top-ups grow — no plan to choose. Your current tier and limit are shown in your account.

Is there an SLA?

The service is in Beta — no formal SLA yet. Live status is at /status, and we report incidents and planned maintenance there and by email.

One model endpoint. Start building.

Connect PRO-1 through the OpenAI-compatible interface your application already understands.

Create an API key View documentation