Inkling-Small
Inkling-SmallAn efficient, open-weight multimodal AI model. Enjoy faster inference, lower costs, and native text, image, and audio support at 1M tokens.
Model details
View repositoryExample usage
Inkling-Small is an efficient, open-weights multimodal model from Thinking Machines Lab that processes text, image, and audio inputs and generates text. Architecturally, it is a sparse mixture-of-experts transformer with 276B total parameters (12B active), a 1M context window, and native vision and audio capabilities.
Designed for exceptional efficiency, it provides a significantly smaller footprint than its predecessor, resulting in faster inference and reduced deployment costs. Like the original Inkling, it is released under open weights and is aimed at developers building agentic workflows, complex RAG applications, and high-performance chatbots that require high reasoning depth at scale.
1# You can use this model with any of the OpenAI clients in any language!
2# Simply change the API Key to get started
3
4from openai import OpenAI
5
6client = OpenAI(
7 api_key="YOUR_API_KEY",
8 base_url="https://inference.baseten.co/v1"
9)
10
11response = client.chat.completions.create(
12 model="thinkingmachines/inkling-small",
13 messages=[
14 {
15 "role": "user",
16 "content": "Implement Hello World in Python"
17 }
18 ],
19 top_p=1,
20 max_tokens=1000,
21 temperature=1,
22 presence_penalty=0,
23 frequency_penalty=0
24)
25print(response.to_json())1{
2 "id": "chatcmpl-xxxx",
3 "choices": [
4 {
5 "finish_reason": "stop",
6 "index": 0,
7 "message": {
8 "content": "Here is the canonical \"Hello, World!\" in Python:\n\n```python\nprint(\"Hello, World!\")\n```\n\n**How it works**\n- `print()` is Python's built-in function for sending output to the console.\n- The text inside the quotes is a string (you can use single `'` or double `\"` quotes).\n\n**Running it**\nSave the code in a file (e.g., `hello.py`) and run:\n```bash\npython hello.py\n```\n\nOutput:\n```\nHello, World!\n```",
9 "role": "assistant",
10 "reasoning_content": "The user wants a \"Hello World\" implementation in Python. This is a very basic request. I should provide the simplest, most standard version, and maybe mention a few common variations (like using print with different styles, or the classic \"Hello, World!\").\n\nThe standard is:\n```python\nprint(\"Hello, World!\")\n```\n\nOr with single quotes:\n```python\nprint('Hello, World!')\n```\n\nI can present this clearly with explanation that `print()` is the built-in function to output text to the console.\n\nSince it's so simple, I should keep it concise but clear. Maybe include how to run it briefly?\n\nLet's provide:\n1. The code\n2. Brief explanation\n3. Maybe execution instructions (optional but helpful for absolute beginners)\n\nActually, the prompt is \"Implement Hello World in Python\" - just need the code.\n\nI'll provide the canonical version with a brief explanation."
11 }
12 }
13 ],
14 "created": 1784140413,
15 "model": "inferact/inkling-small-nvfp4",
16 "object": "chat.completion",
17 "usage": {
18 "completion_tokens": 295,
19 "prompt_tokens": 9,
20 "total_tokens": 304,
21 "completion_tokens_details": {
22 "accepted_prediction_tokens": 0,
23 "audio_tokens": 0,
24 "reasoning_tokens": 182,
25 "rejected_prediction_tokens": 0
26 },
27 "prompt_tokens_details": {
28 "audio_tokens": 0,
29 "cached_tokens": 0
30 }
31 }
32}
