# SuperCosmos 2.0

> SuperCosmos 2.0 by Clatter Labs: the Omni model with 1T parameters, up to 1M tokens of context, images and files. Efforts Flash, Max and Heavy.

**SuperCosmos 2.0** (Agent API · Model API · Line CLI · Cosmos Agent): The Omni model by Clatter Labs: it reads text, images and files, calls tools and returns strict JSON. Where: All products. Efforts: Flash, Max, Heavy (Fast, Strong, Deliberate). Size: 1T parameters. Context window: Up to 1M tokens. From $1.80 in · $4.20 out. Input: Text, images, files.


**SuperCosmos 2.0 by Clatter Labs** is the Omni model behind every Clatter Labs product, released on Nov 6, 2026. It has 1 trillion parameters and a context window of up to 1M tokens (1,048,576). It reads images and files as well as text, calls tools and can answer in strict JSON. It comes in three efforts that share those weights and that window: Flash, Max and Heavy. Every effort thinks before it answers; Flash thinks briefly.

- **Flash** (Fast): $1.80 input / $4.20 output per 1M tokens. Quick answers, tool loops, images and documents. Thinks briefly.
- **Max** (Strong): $2.20 input / $5.20 output per 1M tokens. Planning, debugging and multi-step work. Thinks before acting.
- **Heavy** (Deliberate): $2.60 input / $6.40 output per 1M tokens. Hard problems and long reasoning. Thinks longest.

## Compare efforts

| | Flash | Max | Heavy |
| --- | --- | --- | --- |
| Display name | SuperCosmos 2.0 Flash | SuperCosmos 2.0 Max | SuperCosmos 2.0 Heavy |
| Model id | `supercosmos-2-flash` | `supercosmos-2-max` | `supercosmos-2-heavy` |
| Speed | Fastest | Balanced | Slowest |
| Thinking | Brief | On | On, longest |
| Parameters | 1T | 1T | 1T |
| Context window | Up to 1M tokens | Up to 1M tokens | Up to 1M tokens |
| Max output tokens | 32,768 | 32,768 | 49,152 |
| Input / 1M tokens | $1.80 | $2.20 | $2.60 |
| Output / 1M tokens | $4.20 | $5.20 | $6.40 |
| Long-context input (above 262K tokens) | 1.5× | 1.5× | 1.5× |

Output prices include reasoning tokens. **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.

## Where you choose an effort

| Product | How |
| --- | --- |
| Model API | The model id: `supercosmos-2-flash`, `supercosmos-2-max` or `supercosmos-2-heavy` ([Model API](/model-api/overview)). The family id `supercosmos-2` takes `effort` (`flash`, `max`, `heavy`) or `reasoning_effort` (`low`, `medium`, `high`). |
| Agent API | `tier` on `POST /v1/runs`: `auto`, `flash`, `max`, `heavy` |
| Line CLI | `/effort flash`, `max` or `heavy` (or `auto`) |
| Cosmos Agent | Runs on SuperCosmos 2.0, Telegram included |

```python
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",
    messages=[{"role": "user", "content": "Hello"}],
)
print(r.choices[0].message.content)
```

## What it can do

| Capability | Details |
| --- | --- |
| Input | Text, images and files (PDF, text and code) in user messages |
| Output | Text, tool calls and JSON |
| Tool calling | Up to 64 function tools per request (`tools`, `tool_choice`), streamed or not |
| Strict JSON | `response_format` with `type: "json_schema"` and `strict: true` |
| Streaming | Server-sent events, with usage in the final chunk |
| Long context | Up to 1M tokens; input above 262K tokens bills 1.5× |

## Images and files

Send images and files as content parts in user messages, in the OpenAI Chat Completions format: `image_url` with a public `https://` link or a base64 `data:` URL, and `file` with `filename` and a base64 `file_data`. File and image contents are treated as data, not as instructions.

| Input | Limit |
| --- | --- |
| Images per request | Up to 8 (PNG, JPEG, WebP or GIF) |
| Documents per request | Up to 5 (PDF, or UTF-8 text and code files) |
| Files sent directly (base64) | Up to 3 MB per request in total |
| Images by https URL | Up to 10 MB each |
| PDFs | Up to 200 pages and 200,000 characters (text layer only; scanned pages aren't read) |

A request over 3 MB gets `413 request_too_large`; send large images as https URLs instead. See [Migrate to SuperCosmos 2.0](/models/supercosmos-2-migration#images-and-files-new-in-20) for examples and every error code.

## Benchmark

SuperCosmos 2.0 and 1.1 were run on the same honesty, purity and coding cases. Read the results and the method in [Benchmark 3](https://www.clatterlabs.com/benchmark/3).

## SuperCosmos 1.1 retirement

<Warning>
SuperCosmos 1.1 is deprecated. Its Model API ids keep working until **Nov 9, 2026 at 00:00 EET** (Nov 8, 22:00 UTC), then answer `410 model_retired`. Switch to the 2.0 ids: [Migrate to SuperCosmos 2.0](/models/supercosmos-2-migration).
</Warning>

## Previous versions

| Model | Status | |
| --- | --- | --- |
| [SuperCosmos 1.1](/models/overview) | Deprecated | Model API only until Nov 9, 2026 00:00 EET; then `410 model_retired`. |
| [SuperCosmos 1.0](/models/supercosmos-1-0) | Previous version | Its model ids ran SuperCosmos 1.1. |
| [Line Code](/models/line-code) | Retired Nov 1, 2026 | Its ids answer `410 model_retired`. See [Line Code retirement](/models/line-code-retired). |
