Inkling
NewStandardThinking Machines Lab's open-weight multimodal model with 41B active parameters out of 975B total.
Inkling is built for general-purpose reasoning, coding, agentic and tool-use systems, retrieval-augmented generation, instruction following, and multilingual conversation. Native image and audio understanding with configurable thinking effort and a 524K-token context window.
▼
Inkling is somewhat expensive at $1.00/M tokens input and $4.05/M tokens output. Cached input reads are priced at $0.17/M tokens, making repeated context cheaper. The model supports vision/image input, audio input, extended reasoning, a 524K context window.
Benchmarks
Category Rankings
Specifications
Performance
Providers
Misc
Compare with
Frequently Asked Questions about Inkling▼
When was Inkling released?
▼
Inkling was released on July 17, 2026.
Who created Inkling?
▼
Inkling was created by Thinking Machines Lab.
How much does Inkling cost?
▼
Inkling costs $1.00/M input tokens and $4.05/M output tokens.
What is Inkling API pricing?
▼
The Inkling API is priced at $1.00 per million input tokens and $4.05 per million output tokens. You can access Inkling through idapt.app alongside 200+ other AI models in one workspace.
Is Inkling a reasoning model?
▼
Yes, Inkling is a reasoning model. It can think step-by-step through complex problems before providing an answer, often yielding better results on difficult tasks like math, coding, and multi-step logic.
Does Inkling support image or vision input?
▼
Yes, Inkling supports vision/image input. You can share images in your conversation and the model will analyze and respond based on their content.
Does Inkling support audio?
▼
Yes, Inkling supports audio input, allowing it to process and respond to voice and audio files.
What is the context window of Inkling?
▼
Inkling has a context window of 524K tokens, meaning it can process approximately 393,216 words of context at once.
Is Inkling open source?
▼
Inkling is a proprietary model developed by Thinking Machines Lab. The model weights and training data are not publicly available.
How does Inkling perform on coding?
▼
Inkling resolves 78% of SWE-bench Verified tasks — real GitHub issues fixed end-to-end, the strongest single signal for agentic software engineering.
What is Inkling good at?
▼
Based on benchmark data, Inkling excels at: software development and coding tasks, mathematical problem solving, visual analysis and image understanding. It is Thinking Machines Lab's capable model suitable for a wide range of tasks.
Where can I use Inkling?
▼
You can chat with Inkling on idapt.app — no separate API key required. idapt is an AI workspace with 200+ models, agents, computers, and tasks. You can also access Inkling through Thinking Machines Lab's own API or platform.