# Inkling

Inkling is a generalist. It covers agentic work, reasoning, coding, instruction following, factuality, vision, and audio rather than specializing in one domain, and controllable thinking effort lets you balance answer quality against cost and latency on each request.

- **Model ID:** `thinkingmachines/inkling`
- **Type:** chat
- **Providers:** togetherai, modal, thinkingmachines, deepinfra
- **Context window:** 524,288
- **Maximum output tokens:** 262,144
- **Pricing:** $0.95/1M input tokens, $4.05/1M output tokens
- **Canonical page:** https://vercel.com/ai-gateway/models/inkling

## Supported parameters

Detailed capability metadata has not been reported for this model.

## Example

```ts
import { streamText } from 'ai'

const result = streamText({
  model: 'thinkingmachines/inkling',
  prompt: 'Why is the sky blue?'
})
```

## About

Inkling became available on AI Gateway on July 15, 2026. Inkling is a decoder-only transformer with a sparse Mixture-of-Experts (MoE) backbone: 66 layers, 975 billion total parameters, and 41 billion active per token, with each token routed to 6 of 256 experts plus 2 shared experts. Attention mixes local and global layers, and the context window is 524.3K tokens. Thinking Machines released the weights under the Apache 2.0 license.

Multimodality is native rather than bolted on. Images enter through a hierarchical patch encoder and audio through discrete token encoding, and both are processed jointly with text by the same decoder. Inkling accepts pixel-based images with each dimension between 40px and 4096px, and WAV audio sampled at 16kHz. Inkling transcribes speech, follows spoken instructions, and answers questions about recordings, scoring 91.4% on VoiceBench, 77.2% on MMAU, and 56.6% on Audio MC. On vision, Inkling scores 73.5% on MMMU Pro and 78.1% on CharXiv reasoning questions, rising to 82.0% when it uses a Python tool to zoom into and crop the image.

On agentic and reasoning evaluations at maximum effort, Inkling scores 77.6% on SWE-bench Verified, 54.3% on SWE-bench Pro (public), 63.8% on Terminal-Bench 2.1, 76.0% on MCP Atlas, and 45.5% on Toolathlon Verified. Reasoning results include 87.2% on GPQA Diamond, 97.1% on AIME 2026, and 29.7% on Humanity's Last Exam text-only, which rises to 46.0% with tools. Inkling scores 79.8% on IFBench for instruction following.

Controllable thinking effort is the setting you tune most. Raising effort spends more thinking tokens for higher scores, and lowering it returns answers sooner for less. Inkling reaches a given score at fewer thinking tokens than the open-weights models Thinking Machines compared it against, so sweep the effort setting across a representative slice of your traffic before you fix a value.

Inkling also aims for calibrated confidence. It hedges or says it doesn't know rather than guessing, which helps in forecasting and in any workflow where a confident wrong answer costs more than an uncertain one.

Set the model to `thinkingmachines/inkling` in the AI SDK, Chat Completions API, Responses API, Messages API, or other API formats, from TypeScript or Python. AI Gateway serves Inkling through Together AI, Modal, Thinking Machines, DeepInfra, with retries and failover, and mirrors provider pricing with no markup and no platform fee on inference, including on Bring Your Own Key (BYOK) requests.

## What to consider

Thinking Machines states plainly that Inkling is not the strongest overall model available, open or closed. Its case is breadth: one model that accepts text, images, and audio, follows instructions closely, and exposes an effort dial. Factual recall is the weakest area, with 43.9% on SimpleQA Verified, so pair Inkling with retrieval or search when answers depend on specific facts.

## When to use

### Best For

- **Multimodal Agent Backends:** Text, image, and audio input handled by a single model rather than three
- **Speech and Audio Reasoning:** Transcription, spoken instructions, and questions over longer recordings
- **Chart and Document Vision:** Charts, diagrams, and visual math, with a Python tool for zooming and cropping
- **Tool-Heavy Agent Workflows:** Broad tool use across harnesses, with 76.0% on MCP Atlas
- **Mixed Workload Consolidation:** One generalist covering reasoning, coding, chat, and multimodal input
- **Effort-Tuned Cost Control:** Thinking effort set per request to balance answer quality against latency

### Consider Alternatives When

- **Peak Coding Scores:** Dedicated coding models post higher SWE-bench and Terminal-Bench 2.1 results
- **Factual Recall Workloads:** 43.9% on SimpleQA Verified means knowledge-heavy answers need a retrieval layer
- **Lower Cost Per Task:** Inkling Small matches or beats Inkling on many evaluations at a quarter of the size
- **Text-Only Pipelines:** A text-focused model fits better when image and audio input never apply

## Best for

- **Multimodal Agent Backends:** Text, image, and audio input handled by a single model rather than three
- **Speech and Audio Reasoning:** Transcription, spoken instructions, and questions over longer recordings
- **Chart and Document Vision:** Charts, diagrams, and visual math, with a Python tool for zooming and cropping
- **Tool-Heavy Agent Workflows:** Broad tool use across harnesses, with 76.0% on MCP Atlas
- **Mixed Workload Consolidation:** One generalist covering reasoning, coding, chat, and multimodal input
- **Effort-Tuned Cost Control:** Thinking effort set per request to balance answer quality against latency

## Consider alternatives

- **Peak Coding Scores:** Dedicated coding models post higher SWE-bench and Terminal-Bench 2.1 results
- **Factual Recall Workloads:** 43.9% on SimpleQA Verified means knowledge-heavy answers need a retrieval layer
- **Lower Cost Per Task:** Inkling Small matches or beats Inkling on many evaluations at a quarter of the size
- **Text-Only Pipelines:** A text-focused model fits better when image and audio input never apply

## Frequently asked questions

### What is Inkling designed for?

Broad generalist use. Inkling covers agentic work, reasoning, coding, instruction following, factuality, vision, and audio, rather than optimizing narrowly for one domain. Thinking Machines positions that breadth as the reason to pick it.

### What input types does Inkling accept?

Text, images, and audio, with text output. Images work best with each dimension between 40px and 4096px, and audio should be WAV sampled at 16kHz. All three modalities are processed jointly by the same decoder.

### How does controllable thinking effort work in Inkling?

You set an effort level per request. Higher effort spends more thinking tokens and raises scores on hard tasks; lower effort returns answers sooner. Published benchmark results for Inkling are reported at maximum effort.

### How does Inkling compare to Inkling Small?

Inkling Small is roughly a quarter of the size and matches or beats Inkling on many coding, reasoning, and tool-use evaluations. Inkling stays ahead on world knowledge, scoring 43.9% on SimpleQA Verified against 20.6%, and posts slightly higher audio scores.

### How did Inkling score on agentic coding benchmarks?

77.6% on SWE-bench Verified, 54.3% on SWE-bench Pro (public), and 63.8% on Terminal-Bench 2.1, all at maximum effort. On general agentic evaluations, Inkling scores 76.0% on MCP Atlas and 45.5% on Toolathlon Verified.

### How good is Inkling at audio and vision tasks?

Audio is among its stronger areas, with 91.4% on VoiceBench, 77.2% on MMAU, and 56.6% on Audio MC. On vision, Inkling scores 73.5% on MMMU Pro and 78.1% on CharXiv reasoning questions, or 82.0% when it uses a Python tool to inspect the image.

### What context window does Inkling support?

A context window of 524.3K tokens. That covers long documents, extended tool-calling sessions, and conversations that mix text with images and audio.

### Are the weights for Inkling open?

Yes. Thinking Machines released Inkling with open weights under the Apache 2.0 license, which permits commercial use and modification.

### Does AI Gateway support Zero Data Retention for Inkling?

Yes, Zero Data Retention is available for this model. Zero Data Retention is offered on a per-provider basis. See https://vercel.com/docs/ai-gateway/capabilities/zdr for details.

### How do I call Inkling through AI Gateway?

Set the model to `thinkingmachines/inkling` in the AI SDK, Chat Completions API, Responses API, Messages API, or other API formats, from TypeScript or Python. AI Gateway handles authentication, retries, and failover across `togetherai`, `modal`, `thinkingmachines`, `deepinfra`.

## Links

- [Model page](https://vercel.com/ai-gateway/models/inkling)
- [AI Gateway documentation](https://vercel.com/docs/ai-gateway)
- [Provider model documentation](https://www.together.ai/models/inkling)