H Company Unveils Holo4: Open-Weight Computer-Use Models That Click, Code, and Call Tools Across Desktop, Web, Android, and APIs
H Company has released Holo4, a family of generalist computer-use models for AI agents. A single set of weights can click and type on screens, write code, and call MCP or API tools. Holo4 ships in two sizes: Holo4 27B (dense) and Holo4 35B-A3B (Mixture of Experts, 3B active). Both serve a 256K context window on the H Models API.
Deployability: Apache 2.0 for Commercial Self-Hosting
Holo4 35B-A3B ships under Apache 2.0 weights, enabling commercial self-hosting. Holo4 27B weights are released under CC BY-NC 4.0, so commercial use of the 27B model runs through the H Models API.
What Is Holo4?
Holo4 is a vision-language model purpose-built for computer use. Holo4 27B is fine-tuned from Qwen3.8-27B, while Holo4 35B-A3B is built on Qwen3.6-35B-A3B. Both models pair with H's open hai-agents harness. The harness sends screenshots and tool results to the model, then executes the requested clicks, typing, code, and tool calls.
H Company targets a well-known gap in the agent ecosystem: GUI-only agents fail without a screen, while tool-calling agents stall when an application has no API. Holo4 runs on desktop, web, Android, code sandboxes, and business APIs β the same model, called the same way, on every platform.
Benchmarks: Close to the Frontier at a Fraction of the Cost
According to H Company's benchmark table, Holo4 27B scores 85.2% on OSWorld at $0.08 per task. Its Qwen3.8 27B base scores 84.3% at $0.22 per task. On AndroidWorld, Holo4 27B reaches 85.1%.
Long workflows reveal the remaining gap. On OSWorld 2.0, Holo4 27B scores 61.7% at $1.22 per task. For comparison, Claude Opus 5.5 scores 81.8% at $8.48 per task β a reminder that frontier reliability on extended, multi-step tasks still comes at a premium. Holo4's value proposition is clear: frontier-adjacent performance at roughly an order of magnitude lower cost.
Why It Matters in 2026
As agentic AI moves from demos to production deployments in 2026, model selection increasingly hinges on three factors: licensing flexibility, cross-platform coverage, and cost per completed task. Holo4 addresses all three. The Apache 2.0 release of the 35B-A3B variant gives enterprises a path to self-hosted, commercially usable agents, while the 27B model offers API-based access for teams prioritizing raw benchmark performance.
The dual desktop-plus-API design also reflects a broader 2026 industry shift: rather than forcing a choice between visual computer-use and programmatic tool-calling, the most capable agents now blend both. Holo4's single-model approach β driven through the open hai-agents harness β positions H Company as a contender in the fast-growing computer-use segment, competing directly with proprietary frontier models on cost-efficiency rather than absolute peak performance.
via MarkTechPost