Thinking Machines: Inkling (free)

Inkling is an open-weight multimodal mixture-of-experts model from Thinking Machines Lab. It has 41B active parameters out of 975B total, and it is designed for general-purpose reasoning, coding, agentic and tool-use systems. On api14 it is listed under the model id thinkingmachines/inkling:free, in the text category, and the provider release date is 2026-07-17.

The model accepts text, image and audio as input and produces text as output. Its context window is 1,048,576 tokens, and the maximum output is 262,144 tokens. Because it is a mixture-of-experts design, only a fraction of the total parameters are active per request, which is relevant if you are planning throughput or latency budgets around a large model. The combination of a very large context window and multimodal input means you can send long documents, screenshots, diagrams or audio alongside your instructions in a single request.

Practical uses follow from those capabilities. General reasoning and coding tasks fit the model's stated design goals, so code generation, code explanation and multi-step problem solving are natural fits. Agentic and tool-use systems are also explicitly in scope, which means you can build workflows where the model decides when to call a function and how to chain results. The image and audio inputs open up document understanding, transcription-adjacent pipelines and mixed-media analysis without a separate model for each modality. The long context is useful for repository-scale code review, long meeting or call transcripts, and research over large document sets where you would otherwise need chunking and retrieval.

You call it through api14's OpenAI-compatible API using the model id "thinkingmachines/inkling:free". The supported request parameters are frequency_penalty, include_reasoning, max_tokens, presence_penalty, reasoning, reasoning_effort, seed, stop, temperature, tools and top_p. If you already use an OpenAI-style client, you can point it at api14 and set the model field to this id; the reasoning and reasoning_effort parameters let you control how much internal reasoning the model does, and include_reasoning lets you decide whether that reasoning is returned. The tools parameter is what you use for function calling and agentic loops, and seed helps when you want more reproducible sampling.

One thing to keep in mind is that output is text only, so if you need generated images or audio you will need a different model. The free variant is also worth testing against your own workload before you commit to it, since the page shows a live price table rather than a fixed figure here. If your task is short and simple, a smaller text-only model may be cheaper and faster to run, while Inkling makes more sense when you need the long context, the multimodal inputs or the tool-use behaviour.