Inkling is a general-purpose multimodal model from Thinking Machines that accepts text, image, and audio inputs and generates text outputs. It is described as a model for developers building AI-powered applications, including agentic and tool-use systems, coding assistants, chatbots, and retrieval-augmented generation systems.
Thinking Machines describes Inkling as an efficient open model for text, images, and audio, and says it is their most capable open model with native reasoning over those modalities. The model is released with open weights for research, fine-tuning, and downstream integration.
| Developer | Thinking Machines |
|---|---|
| Model ID | thinkingmachines/inkling |
| Context Window | 1,048,576 tokens (~1.0M) |
| Modalities | Image, Text → Text |
| Capabilities | file input, parallel tool calls, reasoning, streaming, structured (JSON schema) output, tool/function calling, vision |
Multimodal input support
Inkling accepts text, image, and audio inputs and produces text outputs.
Open weights
The model is released with open weights to support research, fine-tuning, and integration into third-party products.
Developer-focused design
Thinking Machines says Inkling is intended for developers building AI-powered applications such as agentic systems, coding assistants, chatbots, and RAG systems.
Reasoning across modalities
Thinking Machines describes Inkling as their most capable open model with native reasoning over text, images, and audio.
General-purpose use
Inkling is presented as suitable for general-purpose conversational use, instruction-following, and other natural language and multimodal tasks.
Agentic applications
Inkling is positioned for agentic and tool-use systems that need multimodal understanding and text generation.
Coding assistants
The model is described as suitable for coding assistants and across multiple coding languages.
Chatbots
Thinking Machines lists chatbots among the application types Inkling is designed to support.
Retrieval-augmented generation
Inkling is presented as appropriate for retrieval-augmented generation systems.
None stated
No attributable comparisons to named alternative models were found in the search results for this exact model id.
Sources: https://huggingface.co/thinkingmachines/Inkling · https://thinkingmachines.ai/inkling/ · https://thinkingmachines.ai/news/introducing-inkling/
Measured on the AI/ML API gateway across live traffic. Figures are medians unless stated. Last updated 2026-10-12.
Inkling is documented as suited to building AI-powered applications that use agents and tools, including agentic and tool-use systems. It is also suited to coding assistants and coding workflows, with the model described as handling coding tasks and writing code to get work done.
It is documented for general-purpose conversational use, instruction-following, and other natural language tasks, including detailed multi-step instructions. It is also suited to multimodal workflows involving text, image, and audio input, including reasoning across those modalities.
import requests r = requests.post( "https://api.aimlapi.com/v1/chat/completions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "thinkingmachines/inkling", "messages": [ { "role": "user", "content": "Hello!" } ] }, ) print(r.json())
const r = await fetch("https://api.aimlapi.com/v1/chat/completions", { method: "POST", headers: { Authorization: `Bearer ${process.env.AIMLAPI_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ "model": "thinkingmachines/inkling", "messages": [ { "role": "user", "content": "Hello!" } ] }), }); console.log(await r.json());
curl -X POST https://api.aimlapi.com/v1/chat/completions \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"thinkingmachines/inkling","messages":[{"role":"user","content":"Hello!"}]}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Cached input |
Inkling is offered at a 50% limited-time discount, and OpenRouter documentation also shows a free tier for thinkingmachines/inkling with 200 requests per day.
| Benchmark | Score | What it measures | Source | Retrieved |
|---|---|---|---|---|
| Intelligence | 25.5 | Composite score across standardised reasoning, knowledge and problem-solving evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Coding | 52.1 | Composite score across standardised coding evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| GPQA Diamond | 87.2% | Google-proof graduate science questions (hardest subset) | Source | September 22, 2026 |
| SWE-bench Verified | 77.6% | Resolving verified real GitHub issues | Source | September 22, 2026 |
| MMAU | 77.2% | Multimodal audio understanding + reasoning | Source | September 22, 2026 |
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
Inkling This page | Agentic workflows and structured output | |||
| Efficient open-weight multimodal reasoning |
Inkling has a 1,048,576 tokens context window.
Inkling takes image, text as input and returns text.
Use thinkingmachines/inkling as the model id. Requests go to https://api.aimlapi.com/v1/chat/completions.
Inkling is priced at input $1.3754 / 1M tokens, output $5.57037 / 1M tokens, cached input $0.233818 / 1M tokens.
Yes, Inkling can stream responses as they are generated.
Yes, Inkling accepts image input alongside text.
Inkling was built by Thinking Machines.
Inkling is a mixture-of-experts model with 41B active parameters out of 975B total.
Inkling supports tools, parallel tool calls, structured output, and reasoning, making it suitable for agentic and tool-use systems.
