Aleph Alpha's open-weight English-German mixture-of-experts model for self-hosting, released under Apache 2.0 in October 2026
Perceptron Mk1.5
New· 4Vision and audio model that takes text, images, video and audio for robots, drones, glasses and phones
About Perceptron Mk1.5
Robots, drones, smart glasses and phones are the devices Perceptron Mk1.5 is designed to help perceive their surroundings. Released by Perceptron on 25 September 2026, it reads text, images, video and audio and replies with text together with structured annotations. The company positions it as the perception and decision layer of a physical agent, one that does not need retraining for each new device. It was announced with a blog post and API release notes on the same date, and it appeared on OpenRouter that afternoon.
Inputs and the annotations it returns
A request can mix text, images, video and audio. The answer can include points, boxes, polygons, time-stamped clips and object tracks that follow a single item through a video. Compared with the earlier Mk1, Perceptron added audio (about 21.8 minutes per item at most), tracking across video, function calling for any tool defined in JSON Schema, a reasoning-effort setting and structured output.
Context size and access
The model card gives a context window of 36,864 tokens shared across all inputs, with 8,192 output tokens; the company blog rounds the window to 32K. It is offered through Perceptron's API and through OpenRouter. The weights are not public; Perceptron will license them commercially if asked. On that date the API also stopped serving two older models, Isaac 0.1 and 0.2, so callers of those must switch.
Speed and accuracy claims
Perceptron reports end-to-end latency two to five times lower than Mk1 on a single H100. In its own egocentric tests it reports 0.94 for locating hands, where the strongest Gemini model it tried scored 0.62. On EgoSchema it trails Gemini 3.8 Flash slightly, at 80.4 against 81.2 per cent. These are the developer's own measurements, and no independent evaluation had been published when the details were last checked on 26 September 2026.
TrustList summary
Perceptron Mk1.5, released on 25 September 2026 by Perceptron, is a model that reads vision and audio. It acts as the perception and decision layer for physical agents such as drones, legged robots, smart glasses and phones, without needing retraining for each device.
What Perceptron Mk1.5 does
- Multimodal input: accepts text, images, video and audio, with audio input of up to about 21.8 minutes per item.
- Structured annotations: returns text plus annotations: point locations, bounding boxes, outlines, timed clips and tracks of an item across a video.
- Function calling: calls any tool described with a JSON Schema and can return structured output.
- Adjustable reasoning: lets the caller set the reasoning effort for each request.
- Changes from Mk1: adds audio input, video tracking and function calling compared with the earlier Mk1 model.
Customers and markets
Teams building physical agents for robots, drones, smart glasses and phones. The weights are not public, and Perceptron offers them under a commercial licence on request.
Key facts
- Released: 25 September 2026
- Context window: 36,864 tokens shared by all inputs, with up to 8,192 output tokens
- Retired models: the older Isaac 0.1 and 0.2 models were retired from the API on the same day
Summary written by TrustList from the business's own published information, Nov 2026.
Benchmarks & AI stats
| Benchmark | Official | Community avg |
|---|---|---|
| No benchmark scores yet. Be the first to add one. | ||
“Official” values are editor-approved and feed the ranking. “Community avg” is the mean of member submissions (shown for transparency; it never affects the ranking until an editor approves a value).
Sign in to add a benchmark score for this model.
Perceptron Mk1.5 launches
Trust Score
An earned signal from verification, reviews, awards, transparency and engagement - the vendor can't buy it.
Updated 9/26/2026
What the business tells TrustList
- Founded
- 2026 (0 years) evidence sent
Request a demo or quote from Perceptron Mk1.5
Reviews
Write the first review of Perceptron Mk1.5
Used it? Your experience helps other buyers decide.
Write a reviewQuestions & answers
No questions yet. Be the first to ask about Perceptron Mk1.5.
Perceptron Mk1.5 alternatives
Similar listings buyers compare against this one.
Open-source 2B decision model from AWS's Strands Agents project that picks options and returns scores with a reliability figure
Open-weight translation model from Bilibili's Index team, covering 150 languages and following glossary and style instructions
Perplexity's open-weight decision model that picks among given options for routing and checks in AI systems
Liquid AI's open-weight decision model, giving yes or no, choice or score answers from text and images
Anthropic's smallest and fastest Claude 5.5 model, available on its own API and three cloud platforms
Microsoft AI's real-time speech-to-text model for voice agents, transcribing 60 languages with continuous language detection
Cantina Security's first open-weight security model, a post-trained GLM-5.3-Flash for vulnerability research and exploit checks
More on TrustList
Everything here links back to the same verified catalogue. Pick your next stop.
- Write a reviewHelp other buyers decide
- CompaniesAgencies, consultancies and IT service providers, ranked by verified reviews.
- ProductsSoftware and SaaS with pricing, features, integrations and alternatives.
- ArticlesGuides, comparisons and research from the editorial desk and the community.
- AwardsAnnual recognition decided by verified reviews and an independent jury.
- LaunchesNew products and releases, voted up by the community every day.
- RequestsBuyers describe what they need; vendors respond directly.
- PeopleReviewers, authors and makers with public profiles.
- ComparePut up to four listings side by side before you shortlist.