Skip to main content
Holo3.1 is our latest family of Vision-Language Models (VLMs) for computer-use agents across web, desktop, and mobile.
ModelActive ParametersMain Use CasesLicenseResources
Holo3.1-35B-A3B3BFast, low-latency computer use across web, desktop, and mobileApache 2.0Model card
Holo3-35B-A3B3BHigh-throughput, low-latencyApache 2.0Model card
Holo3-122B-A10B10BMaximum performance, complex tasksResearch onlyBenchmarks
The open-weight models are on Hugging Face: see each model card above for specs and benchmarks, or browse the full Holo3.1 collection for the other sizes (0.8B, 4B, 9B) and the quantized FP8, GGUF, and NVFP4 builds. Holo3-122B-A10B is API-only, so its weights are not published; the Holo3 blog post covers its specs and performance.

​
Two ways to use Holo

ModePatternOutputWhen to use
Agent loopMulti-turn: conversation + screenshots → next tool call{note, thought, tool_call} or native tool_callsHolo as the brain of an autonomous browser or desktop agent
Element localizationSingle-turn: image + target description → coordinates{x, y} in [0, 1000]UI grounding inside any external agent or pipeline (yours or someone else’s)

​
Get started

1

Get an API key

Generate a key on Portal-H and export it. The free tier gives rate-limited access to holo3-1-35b-a3b, no credit card required.
export HAI_API_KEY="your-api-key-here"
2

Install the OpenAI client

The Models API is OpenAI-compatible, so the official client works as-is, only the base_url changes.
pip install openai
3

Make your first request

Point the client at H by overriding base_url, then send a request. Holo is multimodal: you can send text, images, or both. Here is a minimal text request to confirm your key and client are working.
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.hcompany.ai/v1/",
    api_key=os.environ.get("HAI_API_KEY"),
)

response = client.chat.completions.create(
    model="holo3-1-35b-a3b",
    messages=[{"role": "user", "content": "In one sentence, what is a computer-use agent?"}],
)

print(response.choices[0].message.content)
The same API and code paths work for all models; swap model for holo3-122b-a10b when you need maximum performance.
Holo3 35B is being deprecated in favor of Holo3.1 35B on June 15, 2026. Migrate from holo3-35b-a3b to holo3-1-35b-a3b.
That is the whole setup. To use Holo on real screens, send a screenshot and continue with the agent loop or element localization below.

​
Next steps

Agent loop

How to use Holo in your computer-use harness.

Element localization

Get click coordinates from a screenshot.

API reference

Endpoint, models, parameters, and limits.