Inkling
Neural Network
Inkling is an open multimodal model by Thinking Machines Lab for reasoning, coding, and agentic tasks.
Specifications
- Input
- text, images, documents
- Output
- text
- Context size
- 524 288 tokens
- Max answer length
- 471 859 tokens
- Reasoning effort
- none, minimal, low, medium, high, max
- Capabilities
- reasoning, function calling
Prompt cost
(per 1M tokens)
Answer cost
(per 1M tokens)
How it works Inkling?
Inkling is an open-weights multimodal model by Thinking Machines Lab based on a mixture-of-experts architecture: out of 975 billion parameters, 41 billion are active per step, providing the capacity of a large network with efficient inference. The creators designed it for general reasoning, programming, agentic systems, and tool use, while multimodality allows inputting both text and images. In BotHub, you can access Inkling without a VPN or foreign card: pay with Russian cards in rubles on a pay-as-you-go basis, and tokens do not expire. Other models are available in the same window, making it easy to compare answers and switch without rewriting integrations: access is provided via a unified OpenAI-compatible API, requests are encrypted with AES-GCM, and teams have access to contracts, invoices, electronic document management, and an admin panel with limits. It is useful when you need to analyze and rewrite third-party code, find the cause of a failing test, build an agent with function calling and external tools, extract meaning from diagrams and screenshots, or complete tasks with complex reasoning chains.Frequently asked questions about Inkling
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
It is a text model with multimodal input: it accepts text, documents, and images, and responds with text. It is suitable for analyzing large documents, writing and refactoring code, debugging, log analysis, STEM tasks, and working in agentic scenarios with tool calling.
The output limit is up to 471,859 tokens, and the context window reaches 1,048,576 tokens. This is enough to upload extensive documentation or a codebase and receive a full document, a large code module, or a detailed analysis in response without splitting it into parts.
Yes, the reasoning mode is confirmed: the model builds a chain of steps before the final answer. This significantly helps with math, logic, requirement analysis, and finding the causes of code errors. In BotHub, you pay for tokens as you go; no subscription is needed.
Yes, function calling and structured JSON output are supported. You can connect external tools, databases, and services, build agents, and parse responses without post-processing. Through BotHub's unified OpenAI-compatible API, you can switch models without rewriting your integration.
According to our data, images and documents are accepted as input: interface screenshots, diagrams, charts, tables, PDFs, and long reports. The model extracts data from them and responds with text. For audio and video, BotHub has separate specialized models in the same window.
We do not have specific data on languages, so it is more reliable to test it with your own task: upload a typical document or code snippet and compare the result with other models in the same BotHub window. Switching takes a couple of seconds, and tokens do not expire.