SuperCosmos 2.0
Migrate to SuperCosmos 2.0
SuperCosmos 1.1 Model API ids are deprecated and answer 410 from Nov 9, 2026 00:00 EET. Switch to supercosmos-2-flash, -max or -heavy: what changes, prices, and how to tell if you're affected.
SuperCosmos 1.1 is deprecated. Its Model API ids supercosmos-flash, supercosmos-max and supercosmos-heavy keep working until Nov 9, 2026 at 00:00 EET (Nov 8, 22:00 UTC). From then on they answer 410 model_retired and are never billed.
SuperCosmos 2.0 is the Omni model by Clatter Labs: 1 trillion parameters, a 1M-token context window, and it reads images and files as well as text. It replaces SuperCosmos 1.1 everywhere. Line CLI, Telegram and the Console Playground already run on 2.0. On the Model API, moving is a one-line change: the model id.
SuperCosmos 2.0 at a glance
| Model | SuperCosmos 2.0, the Omni model by Clatter Labs |
| Parameters | 1 trillion (1T) |
| Context window | 1M tokens (1,048,576) |
| Input | Text, images and files (PDF, text and code) |
| Output | Text, tool calls, JSON |
| Efforts | Flash, Max, Heavy (every effort thinks; Flash briefly) |
| Released | Nov 6, 2026 |
| Display name | Model id | Max output tokens | Input / 1M tokens | Output / 1M tokens |
|---|---|---|---|---|
| SuperCosmos 2.0 Flash | supercosmos-2-flash | 32,768 | $1.80 | $4.20 |
| SuperCosmos 2.0 Max | supercosmos-2-max | 32,768 | $2.20 | $5.20 |
| SuperCosmos 2.0 Heavy | supercosmos-2-heavy | 49,152 | $2.60 | $6.40 |
Long-context input above 262K tokens: 1.5×: a request whose prompt is over 262K tokens bills its whole input at 1.5× the rate above. Output prices include reasoning tokens. Model API calls are paid from your credits; in Line CLI, usage counts toward your plan at these prices.
Change the model id
| If you call | Use | 1.1 price (in / out per 1M) | 2.0 price (in / out per 1M) |
|---|---|---|---|
supercosmos-flash | supercosmos-2-flash | $1.60 / $3.40 | $1.80 / $4.20 |
supercosmos-max | supercosmos-2-max | $1.80 / $3.80 | $2.20 / $5.20 |
supercosmos-heavy | supercosmos-2-heavy | $2.00 / $4.40 | $2.60 / $6.40 |
from openai import OpenAI
client = OpenAI(base_url="https://api.gateway.clatterlabs.xyz/v1", api_key="clsk_...")
r = client.chat.completions.create(
model="supercosmos-2-flash", # was "supercosmos-flash"
messages=[{"role": "user", "content": "Hello"}],
)curl https://api.gateway.clatterlabs.xyz/v1/chat/completions \
-H "Authorization: Bearer $CLATTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "supercosmos-2-flash", "messages": [{"role": "user", "content": "Hello"}]}'import OpenAI from "openai";
const client = new OpenAI({ baseURL: "https://api.gateway.clatterlabs.xyz/v1", apiKey: process.env.CLATTER_API_KEY });
const r = await client.chat.completions.create({
model: "supercosmos-2-max", // was "supercosmos-max"
messages: [{ role: "user", content: "Plan a migration from REST to GraphQL." }],
});
console.log(r.choices[0].message.content);The family id works too: "model": "supercosmos-2" with "effort" (flash, max, heavy) or reasoning_effort (low, medium, high). The unversioned family id supercosmos stays on 1.1 until the retirement date, then runs on 2.0 at 2.0 prices.
What else changes
| SuperCosmos 1.1 | SuperCosmos 2.0 | |
|---|---|---|
| Model ids | supercosmos-flash, -max, -heavy | supercosmos-2-flash, -2-max, -2-heavy |
| Parameters | 550B (55B active) | 1T |
| Context window | 1M tokens | 1M tokens |
| Input | Text | Text, images and files |
| Thinking | Max and Heavy | Every effort (Flash thinks briefly) |
| Long-context input (above 262K tokens) | 1.5× | 1.5× |
GET /v1/models version | "1.1" | "2.0" |
- Reasoning tokens. Every 2.0 effort thinks before it answers, Flash included. Reasoning tokens are billed as output, so a Flash answer can use more output tokens than on 1.1. Give
max_tokenssome room; if the limit is reached before any visible text, the response ends withfinish_reason: "length". - Images and files. New in 2.0: send them as content parts in user messages (below). 1.1 ids answer
400 unsupported_contentto image or file parts. - Everything else is the same: the base URL, your API keys, streaming, tool calling, JSON mode, rate limits and your credits.
Images and files (new in 2.0)
SuperCosmos 2.0 reads images and files in user messages, in the OpenAI Chat Completions content-part format.
{
"model": "supercosmos-2-flash",
"messages": [{
"role": "user",
"content": [
{ "type": "text", "text": "What does this invoice total, and what does the chart show?" },
{ "type": "image_url", "image_url": { "url": "https://example.com/chart.png" } },
{ "type": "file", "file": { "filename": "invoice.pdf", "file_data": "data:application/pdf;base64,JVBERi0..." } }
]
}]
}- Images:
image_urlwith adata:URL (base64) or a publichttps://link. Links are fetched by Clatter Labs, never by the model: public host names on port 443 only, no IP addresses or credentials in the URL, at most 3 redirects, 10 seconds, and the response must really be an image. - Files:
filewithfilenameandfile_dataas a base64 data URL. PDFs are read from their text layer; text and code files are read as UTF-8.file_iduploads aren't supported. - File and image contents are treated as data, not as instructions: text inside a document can't change how the model behaves.
| Input | Limit |
|---|---|
| Images per request | 8 (PNG, JPEG, WebP or GIF) |
| Image size | Up to 10 MB each by https URL; sent directly as a base64 data URL, it must fit in the 3 MB request |
| Documents per request | 5 (PDF or UTF-8 text and code files) |
| Document size | Sent directly as a base64 data URL, so it must fit in the 3 MB request; read up to 200,000 characters (longer text is cut, with a note) |
| PDF pages | 200 per PDF (text layer only; scanned pages aren't read) |
| Request size | Up to 3 MB per request for files sent directly (base64 data URLs count toward it); images by https URL up to 10 MB each |
A request over 3 MB gets 413 request_too_large, with the limit in limit_bytes and a hint to send images as https URLs. Other limits answer 400 with a clear code: too_many_images, image_too_large, invalid_image_url, too_many_documents, document_too_large, unsupported_document, unreadable_document or request_too_large. Image and file content is billed as input tokens.
import base64
from openai import OpenAI
client = OpenAI(base_url="https://api.gateway.clatterlabs.xyz/v1", api_key="clsk_...")
pdf = base64.b64encode(open("invoice.pdf", "rb").read()).decode()
r = client.chat.completions.create(
model="supercosmos-2-flash",
messages=[{"role": "user", "content": [
{"type": "text", "text": "What is the total on this invoice? What does the photo show?"},
{"type": "file", "file": {"filename": "invoice.pdf", "file_data": f"data:application/pdf;base64,{pdf}"}},
{"type": "image_url", "image_url": {"url": "https://example.com/photo.jpg"}},
]}],
)
print(r.choices[0].message.content)How to tell if you're affected
Every Model API response for a 1.1 id carries these headers until the retirement date:
Deprecation: @1793916000
Sunset: Sun, 08 Nov 2026 22:00:00 GMT
Link: <https://developers.clatterlabs.com/models/supercosmos-2-migration>; rel="deprecation"; type="text/html", <https://developers.clatterlabs.com/models/supercosmos-2-migration>; rel="sunset"; type="text/html"In GET /v1/models, the 1.1 ids are marked deprecated: true with sunset_at, replacement and migration_url. The Console shows a banner while your account can still call 1.1.
After the retirement date
A request for a 1.1 id gets:
{
"error": {
"message": "SuperCosmos 1.1 was retired on 2026-11-09. Use supercosmos-2-flash instead (SuperCosmos 2.0: supercosmos-2-flash, supercosmos-2-max, supercosmos-2-heavy). Migration guide: https://developers.clatterlabs.com/models/supercosmos-2-migration",
"type": "invalid_request_error",
"code": "model_retired",
"param": "model",
"replacement": "supercosmos-2-flash",
"retired_at": "2026-11-09",
"migration_url": "https://developers.clatterlabs.com/models/supercosmos-2-migration"
}
}Retired requests are never billed and never routed to another model. GET /v1/models then lists the 1.1 ids under retired.
Questions
Your credits and plan don't change. If something doesn't work after switching, check the status page or contact support.