We use cookies. Cookies and browser storage keep you signed in and remember your preferences. With your permission we also measure how the site is used. You can any time. See our cookie policy.

New here? See how it works A 1-minute tour of the playground, models and your API key.

Perceptron Mk1 vs Inkling Small: API price, context & features · api14

Perceptron: Perceptron Mk1 vs Thinking Machines: Inkling Small

A side-by-side comparison of Perceptron: Perceptron Mk1 ($0.52 / 1M tokens) and Thinking Machines: Inkling Small ($1.58 / 1M tokens) using current USD prices and capabilities from the api14 catalog. Both run through the same OpenAI-compatible API.

Specifications

	Perceptron: Perceptron Mk1	Thinking Machines: Inkling Small
Provider	Perceptron	Thinking Machines
Input	$0.52 / 1M tokens	$1.58 / 1M tokens
Output	$5.25 / 1M tokens	$4.20 / 1M tokens
Cache read	-	$0.35 / 1M tokens
Context window	32,768	524,288
Max output	8,192	262,144
Input	text, image, video	text, image, audio
Output	text	text
Tool calling
Structured output
Reasoning control
Added	May 12, 2026	Jul 30, 2026

Key differences

Perceptron: Perceptron Mk1 is 67% cheaper on input tokens.
Thinking Machines: Inkling Small is 20% cheaper on output tokens.
Thinking Machines: Inkling Small reads 16.0× more context (524,288 tokens).
Only Thinking Machines: Inkling Small supports tool calling.
Only Perceptron: Perceptron Mk1 supports structured JSON output.
A request with 1,000 input and 500 output tokens costs $0.00315 with Perceptron: Perceptron Mk1 and $0.00368 with Thinking Machines: Inkling Small.

In depth

Perceptron: Perceptron Mk1 is a text model from Perceptron, released on 12 May 2026. Thinking Machines: Inkling Small is a text model from Thinking Machines, released later, on 30 July 2026. Both are served through api14 and both take multimodal input and return text, so the choice between them comes down to how much context you need, what kinds of media you are feeding in, and which request parameters you rely on.

The most obvious split is context window. Inkling Small supports a context of 524288 tokens, which is much larger than the 32768 tokens Perceptron Mk1 allows. On input, Perceptron Mk1 accepts text, images and video, while Inkling Small accepts text, images and audio. So video input points to Perceptron Mk1 and audio input points to Inkling Small. Both output only text. They differ on parameters as well: Perceptron Mk1 is the one that exposes structured outputs, which matters if you need the response to conform to a schema. Inkling Small is the one that supports tools, so it is the side to pick for function calling and tool-driven workflows. Perceptron Mk1 also lists frequency_penalty, include_reasoning, max_tokens, presence_penalty, reasoning, temperature, top_k and top_p. Inkling Small lists those plus logit_bias, min_p, reasoning_effort, repetition_penalty, seed, stop, tool_choice and tools. Reasoning support is not a meaningful differentiator here.

For practical selection, reach for Perceptron Mk1 when your task turns on structured output, when you need to send video alongside text, or when you want the model that is cheaper on input. Reach for Inkling Small when the job involves long documents that will not fit in 32768 tokens, when you need audio input, when you need tool or function calling, or when you want the model that is cheaper on output. If your workload is a mix, the context limit is usually the first thing to check: anything large will not fit in Perceptron Mk1 at all, so the decision is made for you. If context fits comfortably and you need a schema-constrained response, the structured outputs support on Perceptron Mk1 makes it the simpler option.

Calling either model through api14 uses the same OpenAI-compatible surface. You point your client at the api14 base URL, send your request to the chat completions endpoint, and set the model field to the id you want. The only change between the two is that string: use perceptron/perceptron-mk1 for Perceptron Mk1 and thinkingmachines/inkling-small for Inkling Small. Your surrounding code, your message formatting and your response parsing stay the same as long as you only use parameters the chosen model supports, so it is worth checking that list before you move a request across. Keep in mind that tool_choice and tools only exist on Inkling Small; structured_outputs only exists on Perceptron Mk1. The live price table on this page shows the current rates for each, so compare input and output pricing there before committing.

There are cases where the other model is still the better pick. If you are already using Perceptron Mk1 for its structured outputs but your prompt grows past its context limit, Inkling Small is the alternative even though it lacks that parameter. Conversely, if you need tool calling in Inkling Small but your workflow also requires a strict output schema, you will have to handle the shaping on your side, or keep that particular call on Perceptron Mk1.

Switch with one string

Both models use the same endpoint and request shape. Change the model value from "perceptron/perceptron-mk1" to "thinkingmachines/inkling-small" and prompts, streaming and tools keep working.

Perceptron: Perceptron Mk1 details & playground →Thinking Machines: Inkling Small details & playground →

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api14.com/api/v1",
  apiKey: process.env.API14_API_KEY,
});

const result = await client.chat.completions.create({
  "model": "thinkingmachines/inkling-small",
  "messages": [
    {
      "role": "user",
      "content": "Explain vector databases in two sentences."
    }
  ]
});
console.log(result.choices[0].message.content);

Perceptron: Perceptron Mk1 vs Thinking Machines: Inkling Small

Specifications

Key differences

In depth

Switch with one string

More comparisons