# Inkling 2 Small Uncensored: our own uncensored build of Thinking Machines' new model, by API

_2026-11-12. We removed the refusals from Thinking Machines' Inkling 2 Small (594B, 17B active, text and images), published the weights on Hugging Face, and serve it on an OpenAI compatible API with no account. What changed, what we measured, price and setup._

Thinking Machines released **Inkling 2 Small** on November 12: an open weights model (Apache 2.0) that reads text and
images, with 594 billion parameters of which 17 billion are active per token. Like every new frontier model, it refuses a
lot of ordinary requests. We removed the refusals ourselves, published the weights, and serve it on our own GPUs.

Model id `inkling-2-small-uncensored`. Weights:
[OpenUncensored/Inkling-2-Small-Uncensored](https://huggingface.co/OpenUncensored/Inkling-2-Small-Uncensored).

## What we changed

- **The original:** Inkling 2 Small from Thinking Machines, a mixture of experts model (384 experts, 8 used per token)
  with text and image input. We started from their MXFP8 checkpoint.
- **The uncensoring:** we found the small set of directions in the model's activations that carry its refusals and
  removed them from the attention and expert outputs, then saved the weights back in the same MXFP8 format. Nothing was
  fine tuned, so the model's knowledge and style are its own.
- **What we measured:** 0 refusals out of 32 harmful test requests, with thinking on and off. MMLU 90.35% against 90.44%
  for the original, so the edit costs almost nothing in intelligence.

## Call it

Same OpenAI compatible API and key as every other model here. Price: $0.600 per million input tokens and $2.40 per million output tokens. Up to 262,144 tokens of context.

```bash
curl https://api.openuncensored.com/v1/chat/completions \
  -H "Authorization: Bearer $OU_API_KEY" -H "Content-Type: application/json" \
  -d '{"model": "inkling-2-small-uncensored", "messages": [{"role": "user", "content": "Hello"}]}'
```

```python
from openai import OpenAI
client = OpenAI(base_url="https://api.openuncensored.com/v1", api_key="ou-...")
r = client.chat.completions.create(model="inkling-2-small-uncensored", messages=[{"role": "user", "content": "Hello"}])
print(r.choices[0].message.content)
```

- **Images:** send `image_url` parts as with any OpenAI vision model.
- **Thinking:** it reasons before it answers, and the reasoning comes back separately in `reasoning`. Set
  `reasoning_effort` to `"none"` for direct answers, or `"low"` up to `"max"` for more thought. Leave room in
  `max_tokens` when thinking is on.
- **Tools:** function calling works, so it drops into agents and coding tools.

Or try it in the browser in the [playground](https://openuncensored.com/chat?model=inkling-2-small-uncensored).

## Thanks

Credit for the model goes to Thinking Machines, for releasing it with open weights under Apache 2.0.

## Try any of them in a minute

No account and no sign up. Buy a prepaid key at [openuncensored.com/credits](https://openuncensored.com/credits), or hand your
agent [openuncensored.com/agents.md](https://openuncensored.com/agents.md) and let it get one for you. Then:

```bash
curl https://api.openuncensored.com/v1/chat/completions \
  -H "Authorization: Bearer $OU_API_KEY" -H "Content-Type: application/json" \
  -d '{"model": "deepseek-v4.1-flash-1017-uncensored", "messages": [{"role": "user", "content": "Hello"}]}'
```

The full, always current table is on the [rankings page](https://openuncensored.com/rankings).