Inkling from Thinking Machines is now available on AI Gateway.

Inkling is a broad generalist model, trained across agentic, reasoning, coding, instruction-following, factuality, vision, and audio tasks rather than optimized for a single domain. The model also supports controllable thinking effort.

To use Inkling, set model to thinkingmachines/inkling in the AI SDK:

import { streamText } from 'ai';

const result = streamText({

model: 'thinkingmachines/inkling',

prompt: 'Summarize this report and list the key risks.',

});

AI Gateway provides a unified API for calling models, tracking usage and cost, and configuring retries, failover, and performance optimizations for higher-than-provider uptime. It includes built-in custom reporting, Zero Data Retention support, budgets for API keys, routing rules, and more.

AI Gateway reflects provider pricing with no markup and does not charge a platform fee on inference, including on Bring Your Own Key (BYOK) requests.

Try Inkling in the model playground.

AI Gateway: Track top AI models by usage

The AI Gateway model leaderboard tracks the most popular models over time, ranking them by the total volume of tokens processed across all Gateway traffic.

View the leaderboard