Perceptron Mk1.5 is Perceptron’s embodied reasoning model for physical agents. It accepts text, image, video, and audio input, and returns text with optional structured annotations such as points, boxes, polygons, tracks, and clips.
The model is described as Perceptron’s most performant public model and adds native audio support, video tracking, web search, sub-agent calls, and more complex visual reasoning.
| Developer | Perceptron |
|---|---|
| Model ID | perceptron/perceptron-mk1.5 |
| Release Date | September 25, 2026 |
| Context Window | 36,864 tokens (~37K) |
| Max Output Tokens | 8,192 tokens |
| Modalities | Image, Text, Video → Text |
| Capabilities | reasoning, streaming, structured (JSON schema) output, tool/function calling, vision |
Multimodal input and structured output
Mk1.5 accepts text, images, video, and audio, and can answer with text plus structured annotations including points, boxes, polygons, tracks, and clips.
Tool calling and workflows
The model supports arbitrary tools defined in the OpenAI format with JSON Schema, and it can decide when to call them.
Grounded reasoning across space and time
Perceptron describes Mk1.5 as understanding images, video, and audio while grounding answers in space and time, including object tracking and locating events over time.
Streaming and structured outputs
Perceptron’s documentation and model listings state that Mk1.5 supports use in chat-completions workflows for questions, transcription, structured extraction, object tracking, and tool-assisted workflows, with structured outputs via JSON Schema.
Physical-agent focus
The model is positioned for embodied or physical-agent use cases, including visual reasoning and sensor-rich interaction.
Image, video, and audio understanding
Mk1.5 is presented as suitable for questions, transcription, structured extraction, and multimodal understanding across images, video, and audio.
Object tracking and spatial localization
The model can emit points, boxes, polygons, clips, and object tracks, making it useful for localization and tracking tasks in images and video.
Tool-assisted workflows
Mk1.5 can call application-provided functions and support workflows that combine model reasoning with external tools such as custom functions or web search.
Embodied-agent applications
Perceptron positions Mk1.5 for embodied agents and physical-agent tasks that require grounded perception and action-oriented reasoning.
Perceptron Mk1
OpenRouter’s comparison page identifies Mk1.5 as the newer model relative to Perceptron Mk1.
Sources: https://www.perceptron.inc/blog/introducing-perceptron-mk1-5 · https://docs.perceptron.inc/perceptron-mk1.5/index.md · https://x.com/perceptroninc/status/2103508527813669193
Measured on the AI/ML API gateway across live traffic. Figures are medians unless stated. Last updated 2026-10-12.
Autonomous drone missions are a documented use case for Mk1.5, including short missions for home inspection, construction site mapping, and search and rescue. It is described as reasoning about what it sees in real time during these flights.
Smart glasses assistance is another documented use case. Mk1.5 is said to power smart glasses and help with navigation by using OCR and web search to supplement the wearer’s surroundings.
Robotic quadruped control is also documented. The model is described as controlling quadrupeds and as being deployed on robotic dogs.
Multimodal question answering and analysis is documented as a general use case. The model is presented as supporting image, video, and audio understanding, including questions, transcription, object tracking, and locating events in time.
import requests r = requests.post( "https://api.aimlapi.com/v1/chat/completions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "perceptron/perceptron-mk1.5", "messages": [ { "role": "user", "content": "Hello!" } ] }, ) print(r.json())
const r = await fetch("https://api.aimlapi.com/v1/chat/completions", { method: "POST", headers: { Authorization: `Bearer ${process.env.AIMLAPI_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ "model": "perceptron/perceptron-mk1.5", "messages": [ { "role": "user", "content": "Hello!" } ] }), }); console.log(await r.json());
curl -X POST https://api.aimlapi.com/v1/chat/completions \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"perceptron/perceptron-mk1.5","messages":[{"role":"user","content":"Hello!"}]}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Benchmark | Score | What it measures | Source | Retrieved |
|---|
The primary published source is Perceptron’s blog post “Introducing Perceptron Mk1.5.” Source: https://www.perceptron.inc/blog/introducing-perceptron-mk1-5
Perceptron Mk1.5 accepts text, images, video, and audio input.
It returns text plus optional structured annotations such as points, boxes, polygons, tracks, and clips.
Yes. It understands images and video, and it is described as supporting more complex visual reasoning.
Yes. Native audio support is listed as a new capability, and the model can understand audio input.
Yes. It can track objects in video and emit object tracks as timestamped geometries.
Yes. It supports arbitrary tools and can call functions declared in the OpenAI format with a JSON Schema.
Yes. It supports structured outputs via JSON Schema.
Yes. Streaming is listed among its supported capabilities.
Yes. Web search is listed among its new capabilities.
Yes. Sub-agent calls are listed as a capability, and the model can dispatch its own sub-agents.
