The release includes Holo4-27B, a 27B dense model, and Holo4-35B-A3B, a 35B mixture-of-experts model. It also updates Holotron4 Nano.
Hcompany says the models can operate through GUIs, code, MCP, and APIs, and can be used across desktops, the web, Android, code sandboxes, and business APIs. Both Holo4 sizes are available on the H Models API. The company has also published weights on Hugging Face in BF16, FP8, NVFP4, and 4-bit GGUF formats, plus benchmark trajectories.
On OSWorld 2.0, Hcompany reports a 61.7% score for Holo4-27B and 30.9% for Holo4-35B-A3B. It says its task factory has produced about 10,000 tasks spanning web apps, MCP servers, and desktop environments. The source does not establish independent replication of these scores, comparable harness conditions for all cited models, licensing terms, or production reliability.