Skip to content
Klyroe
PricingDownloadStart free

Klyroe · Models

Thinking, composed.

Five language models and one for images. Each is tuned for a different part of the work, from a quick answer to a long run of careful steps, so every task gets the right voice.

Try them in KlyroeMeet the lineup

The lineup

Named after musical forms, from a short prelude to a full symphony. Start with Suite 1; switch when the work asks for something else.

Dots compare the models with each other, not with other companies.

  1. Prelude 1

    Fastest

    Quick answers and everyday code at the lowest price.

    Reasoning
    Speed
    Long tasks

    A prelude is short and opens the way.

    Good for

    • High-volume requests
    • Latency-sensitive features
    • Short answers, lookups and simple code

    Less suited to

    • Hard, multi-step problems
    • Reading images

    Try asking

    • Turn this list into a table.
    • What does this error mean?
    • Write a regex for UK postcodes.
    Context
    128K (32K in the app today)
    Max output
    32K
    Thinking
    Low or high
    Tools
    Yes
    Vision
    No
    Input
    $0.10 / 1M
    Output
    $0.50 / 1M

    Proposed API list prices.

    Use it in KlyroeAPI docs

  2. Étude 1

    Light

    Small, quick edits: renames, formatting, short fixes and simple questions.

    Reasoning
    Speed
    Long tasks

    An étude is a short piece that practises one thing well.

    Good for

    • Renames and formatting
    • Short, obvious fixes
    • Simple questions

    Less suited to

    • Long agent runs
    • Decisions with many trade-offs

    Try asking

    • Rename userId to accountId everywhere in this file.
    • Fix the indentation.
    • Make this sentence shorter.
    Context
    128K (32K in the app today)
    Max output
    32K
    Thinking
    Yes
    Tools
    Yes
    Vision
    No
    Input
    $0.12 / 1M
    Output
    $0.60 / 1M

    Proposed API list prices.

    Use it in KlyroeAPI docs

  3. Suite 1

    Default

    Well-rounded and the one to start with. Code, writing, plans, and it reads images.

    Reasoning
    Speed
    Long tasks

    A suite is a set of movements that belong together.

    Good for

    • Getting started, or when you're not sure
    • Writing and fixing code
    • Explaining, planning and reading screenshots

    Less suited to

    • The very hardest problems (try Symphony 1)
    • The lowest possible price per request (try Prelude 1)

    Try asking

    • Here's a screenshot of the bug. What's wrong?
    • Plan a launch email sequence for this product.
    • Explain this function line by line.
    Context
    128K (32K in the app today)
    Max output
    32K
    Thinking
    Yes
    Tools
    Yes
    Vision
    Yes
    Input
    $0.25 / 1M
    Output
    $1.50 / 1M

    Proposed API list prices.

    Use it in KlyroeAPI docs

  4. Fugue 1

    For agents

    Built for coding agents: long runs of tool calls, file edits and shell commands without losing the thread.

    Reasoning
    Speed
    Long tasks

    In a fugue, voices enter one after another and keep the thread.

    Good for

    • Agents that edit a whole project
    • Long chains of tool calls
    • Work that reads far more than it writes (cheap cached input)

    Less suited to

    • One-off hard questions (try Symphony 1)
    • Reading images

    Try asking

    • Add a settings page and wire it to the API.
    • Find every place we parse dates and make them consistent.
    • Run the tests and fix what fails.
    Context
    128K (32K in the app today)
    Max output
    32K
    Thinking
    Yes
    Tools
    Yes
    Vision
    No
    Input
    $0.20 / 1M
    Output
    $0.90 / 1M

    Proposed API list prices.

    Use it in KlyroeAPI docs

  5. Symphony 1

    Most capable

    Our strongest reasoning, for hard bugs, architecture and long careful answers. In Code and agents it uses a faster helper model for routine steps like reading files, and in chat for quick voice replies; you can turn that off in Settings.

    Reasoning
    Speed
    Long tasks

    A symphony brings every section in for the largest work.

    Good for

    • Hard bugs
    • Architecture and design decisions
    • Long, careful answers that should be right first time

    Less suited to

    • Quick replies (it takes longer to think)
    • High volume on a budget
    • Reading images

    Try asking

    • Why does this only fail under load?
    • Compare these two database designs for our case.
    • Review this plan and find what's missing.
    Context
    128K (32K in the app today)
    Max output
    32K
    Thinking
    Yes
    Tools
    Yes
    Vision
    No
    Speed
    Moderate
    Input
    $0.75 / 1M
    Output
    $3.75 / 1M

    Proposed API list prices.

    Use it in KlyroeAPI docs

And for pictures

  1. Canvas 1

    Images

    Makes and edits images from a text description: illustrations, mockups, icons and photos for slides.

    The image model: from a sentence to a picture.

    Good for

    • Illustrations and icons
    • Mockups and slide pictures
    • Editing an image you already have

    Less suited to

    • Long or exact text inside an image
    • Charts that must match real numbers (make those in a sheet or dashboard)

    Try asking

    • A watercolour of a lighthouse at dawn, square.
    • Make the sky in this photo stormy.
    • A flat icon of a paper plane.
    Does
    Text to image, image editing
    Up to 1 MP
    $0.01 / image
    Up to 2 MP
    $0.02 / image

    Proposed API list prices.

    Use it in KlyroeAPI docs

The models run in the cloud through third-party model providers, in the Klyroe app and through the Klyroe API. Symphony runs on open models we select and tune.

The Klyroe API

The API is open: create a free test key in the developer console. Live keys are paid from prepaid API credit, which can't be bought online yet; email hello@senovasys.com to add some. The prices below are proposed list prices and may change; the API docs show what a request costs today. See the proposed subscription plans too.

Compare the language models

ModelBest atContext (hosted)Context (app today)Max outputSpeedVisionToolsThinking
Suite 1Everyday default; code, writing, images in128K32K32KFastYesYesYes
Symphony 1Hardest problems, careful reasoning128K32K32KModerateNoYesYes
Fugue 1Coding agents, long tool use128K32K32KFastNoYesYes
Prelude 1Quick answers, high volume128K32K32KFastestNoYesLow / high
Étude 1Simple edits, short tasks128K32K32KFastNoYesYes

Your answers come from the model you pick. Background tasks, such as summarising long conversations, saving memories, suggesting an agent and the voice greeting, use Prelude, a smaller and faster Klyroe model, whatever model you pick. This can't be turned off.

API pricing

Prices are per million tokens. Cached input applies when a request starts with the same text as a recent request, such as a long system prompt or the files an agent already read. Thinking tokens are billed as output.

ModelInputCached inputOutputBatch (50% off)Example: 1,000 chats*
Suite 1$0.25$0.05$1.50$0.125 / $0.75$1.80
Symphony 1$0.75$0.15$3.75$0.375 / $1.875$4.88
Fugue 1$0.20$0.05$0.90$0.10 / $0.45$1.23
Prelude 1$0.10$0.03$0.50$0.05 / $0.25$0.65
Étude 1$0.12$0.03$0.60$0.06 / $0.30$0.78
Canvas 1$0.01 / image up to 1 MP$0.02 up to 2 MP$0.005 / $0.01$10.00 (1,000 images)

* One chat = 3,000 input tokens and 700 output tokens, no caching. A typical coding-agent task reads far more than it writes, so cached input matters most there: a 25-step task on Fugue 1 with 300,000 input tokens (90% cached) and 12,000 output tokens costs about $0.03.

Speed labels are relative. On the hosted API we expect the fastest models to stream well over 100 tokens per second per request; we will publish measured figures before launch.

Voice API pricing

Bring Klyroe's voices to your own app with an API key: text to speech with Aria and speech to text with Notation, OpenAI-compatible, paid from the same prepaid credit. See the voice API docs.

ModelDoesPriceExample
Aria 1Text to speech, 28 English voices$6 per 1M characters$0.006 for 1,000 characters
Aria 2Text to speech, 23 languages$10 per 1M characters$0.01 for 1,000 characters
Notation 1Speech to text$0.003 per minute, billed per second$0.18 for an hour
Notation 2Speech to text, more accurate, 30 languages$0.006 per minute, billed per second$0.36 for an hour

Aria 2 and Notation 2 are switched on as they launch; GET /v1/audio/voices lists what your key can call today.

Which one should I use?

If you want toUseWhy
Get started, or you are not sureSuite 1Good at almost everything, reads images, quick.
Run an agent that edits a projectFugue 1Holds up over long chains of tool calls; cheap input.
Solve something hard onceSymphony 1Spends longer thinking and gets more right first time.
Handle lots of short requestsPrelude 1Lowest price and fastest replies.
Make a small, obvious editÉtude 1Light and quick for simple changes.
Make a pictureCanvas 1Text to image and image editing.

Klyroe models are built on open-source models used under their licences; see Open-source licenses in the Klyroe app under Settings › About. Prices in US dollars, excluding taxes. Proposed prices dated 27 October 2026.