DEEN
Werk 1.1Ottili

Ottili Werk 1.1

More than a coding model.

Not an assistant that waits for the next prompt. An agent: it plans, uses tools and finishes the job. The model behind Alran Code, and yours through one API.

225tokens per second, up to
$0.50per million input tokens
$1.50per million output tokens
256Ktokens in a single answer

Not an assistant. An agent.

A task is dozens of calls, not one. Werk is tuned for the loop an agent runs: think, call a tool, read what came back, decide. Then again, until the job is done.

  1. Plan: booking page, calendar, save, email, tests
  2. Read (src/pages/listing.mjs)
  3. Wrote (src/booking/slots.mjs)
  4. Wrote (db/migrations/004_bookings.sql)
  5. Ran Apply the bookings migration
  6. Ran Run all tests · 24 passed
  7. Done: every listing can be booked online.

Fast

Every token per second adds up.

An agent writes plans, code, commands and answers, call after call. Speed is the difference between watching it work and getting on with your day.

Time to write 40,000 tokens, about one feature’s worth of output

225 tok/s Werk 1.1, up to2:58
100 tokens per second6:40
50 tokens per second13:20

Pure generation time at a constant rate. Real tasks add reading, tool runs and your reviews.

See it write.

Speed

225tok/s

Output speed, up to. An agent calls the model dozens of times per task; every second adds up.

ottili/werk-1.1 · a customer email0 tokens · 0.0 s
Subject: Your viewing on Friday, 10:30

Hi Ms. Weber,

thanks for booking a viewing of the flat on Lindenstraße 4.
You're confirmed for Friday at 10:30. Here is what to bring:

  - a photo ID
  - proof of income for the last three months
  - any questions about the heating and the lease

If the time no longer suits you, reply to this email or move it
in your booking link. We'll hold the slot until Thursday, 18:00.

See you on Friday,
Fernhill Homes
225 tokens per second.

01

Reasoning, tool use and agentic work

Thinks in steps, calls tools, recovers when a step fails.

02

Coding

Reads the codebase, writes the change, runs the tests.

03

Business tasks

Emails, documents, tables and research. The work around the code.

$0.50

per million input tokens

$1.50

per million output tokens

$0.03

per million cached tokens

Tuned for three kinds of work.

The code, the tools around it and the business work in between. One model for all of it.

Coding

Reads the codebase. Writes the change.

Refactors, features, bug fixes and the tests that prove them, across many files at once.

Tool use and agents

Calls tools. Recovers when one fails.

Structured tool calls, long multi-step runs, and a plan it keeps to when a step goes wrong.

Business tasks

The work around the code.

Emails, documents, tables and research, in the tone and format you need.

Affordable

Same task. A fraction of the price.

Pay per token, nothing else. No seats, no base fee. The welcome credit you get on sign-up covers Werk.

A quick fix80K tokens in · 8K out

Ottili Werk 1.1$0.05
Claude Sonnet 5.5$0.24
Claude Opus 5.5$0.48

4.6× less than Claude Sonnet 5.5 for the same tokens

A new feature800K tokens in · 60K out

Ottili Werk 1.1$0.49
Claude Sonnet 5.5$2.20
Claude Opus 5.5$4.40

4.5× less than Claude Sonnet 5.5 for the same tokens

A big refactor4M tokens in · 300K out

Ottili Werk 1.1$2.45
Claude Sonnet 5.5$11
Claude Opus 5.5$22

4.5× less than Claude Sonnet 5.5 for the same tokens

Per task at list price per million tokens, without cache savings (cached input on Werk costs $0.03). The token counts are examples; alran usage shows yours.

On the Ottili AI Platform

Already on the OpenAI SDK? Change two lines.

Point the base URL at the platform and set the model. Chat Completions, Responses and Anthropic Messages under one key.

  • POST /v1/chat/completions
  • POST /v1/responses
  • POST /v1/messages

Python

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://ai.ottili.one/v1",
    api_key=os.environ["OTTILI_API_KEY"],
)
reply = client.chat.completions.create(
    model="ottili/werk-1.1",
    messages=[{"role": "user", "content": "Summarize this ticket."}],
)

TypeScript

import OpenAI from 'openai';

const client = new OpenAI({
  baseURL: 'https://ai.ottili.one/v1',
  apiKey: process.env.OTTILI_API_KEY,
});
const reply = await client.chat.completions.create({
  model: 'ottili/werk-1.1',
  messages: [{ role: 'user', content: 'Summarize this ticket.' }],
});

curl

curl https://ai.ottili.one/v1/chat/completions \
  -H "Authorization: Bearer $OTTILI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "ottili/werk-1.1",
       "messages": [{"role": "user", "content": "Summarize this ticket."}]}'

Tech specs.

Everything in one table.

Ottili Werk 1.1 technical specifications
Model idottili/werk-1.1
Output speedUp to 225 tokens per second
Context window1,000,000 tokens
Longest answer256,000 tokens
CapabilitiesStreaming, tool calls, reasoning with adjustable effort
InputText
APIOpenAI Chat Completions and Responses, Anthropic Messages
Price per 1M tokens$0.50 input · $0.03 cached input · $1.50 output
Processed inThe United States

Questions.

Is Werk only for code?

No. It is tuned for agentic work in general: coding, tool use and business tasks like emails, documents, tables and research.

Can I use Werk without Alran Code?

Yes. Create an API key on the Ottili AI Platform and call it from any OpenAI- or Anthropic-compatible client.

What does the welcome credit cover?

Werk 1.1. Claude Sonnet 5.5 and Opus 5.5 run on your own Anthropic key.

Where is my data processed?

Werk 1.1 is processed in the United States.

Can it read images?

Werk 1.1 takes text. For images, pick a model that accepts them, for example your own provider.

Put Werk to work.

In Alran Code it is already the default. In your product it is one API key away.