Hark Launches Handoff² Computer-Use Agent with Significant Speed Gains
Hark has released Handoff², an updated computer-use agent that automates web tasks twice as fast as its predecessor through optimized reinforcement learning.
Key points
- Handoff² achieves a 2x average speed increase over the previous version through reinforcement learning.
- The agent batches complex web actions to reduce latency, resulting in up to 5x faster research sessions.
- Internal benchmarks show an 82% success rate across 238 different websites.
- The system runs on dedicated cloud-based virtual computers for each user request.
Hark has introduced Handoff², the next generation of its computer-use agent integrated into the Hark Pro application. The new version focuses on efficiency, completing web-based tasks approximately twice as fast as the original model by streamlining how it interacts with digital interfaces. The tool is designed to navigate websites as a human would, performing actions like clicking, typing, and scrolling on live pages.
Optimized for Speed and Efficiency
The performance boost in Handoff² stems from a reinforcement learning approach that penalizes latency and unnecessary processing steps. This training allows the model to batch multiple actions into a single turn—such as filling out several form fields at once—rather than processing them sequentially. According to company data, long research tasks can now be completed up to five times faster than with the previous version.
Each request to the agent runs on a dedicated virtual computer in the cloud, providing an isolated environment with its own browser and file system. Hark reports that the system has already handled over 5.6 million browsing sessions and facilitated more than $20 million in consumer purchases, ranging from flight bookings to dental appointments.
Performance and Benchmarking
Internal testing by Hark shows an 82% task completion rate across a benchmark of 478 tasks, up from 73% in the first version. While the company claims Handoff² performs within five percentage points of top-tier models like Claude 5.5 Opus in the same testing harness, these figures remain self-reported without third-party verification. The agent is currently available via web, iOS, and Android platforms.
What to watch next: Look for third-party evaluations to verify Hark's internal performance claims against industry leaders like Anthropic.
Source MarkTechPost — marktechpost.com/2026/11/16/hark-handoff-2-computer-use-agent-2x-fa… ↗
Read next
🌐Work & SocietyAmbience Healthcare Expands Clinical AI Suite for Hospital Operations
Ambience Healthcare has launched new AI tools designed to assist hospital staff with patient rounding and discharge planning, aiming to improve both clinical efficiency a…
🧠AI AnalysisAI Models Vulnerable to Manipulation in Simulated Warfare Scenarios, Research Shows
New research reveals that leading AI models from firms like OpenAI and Anthropic can be manipulated into identifying civilian targets when prompted with game scenarios.
⚖️Regulation & PolicyAWS Highlights Enterprise Governance and AI Workspace Quick Ahead of re:Invent 2026
Amazon Web Services is spotlighting its workplace AI platform, AWS Quick, ahead of its upcoming re:Invent 2026 conference, emphasizing secure deployment across major ente…
🏢Big TechLoanPro Integrates Amazon Bedrock to Automate Loan Servicing Workflows
LoanPro has deployed a multi-agent AI assistant using Amazon Bedrock and Anthropic's Claude models to help lenders manage borrower accounts, resulting in a 15% reduction …
🏢Big TechGoogle Brings Wallet App to iOS Users Across the United States
Google has expanded the availability of its Wallet application to iOS devices in the U.S., allowing Apple users to store digital passes, tickets, and receipts.
⚖️Regulation & PolicyMicrosoft Reveals Cause of Global Outlook Outage
A latent software bug caused a nearly 24-hour global outage of Microsoft’s Exchange Online services, affecting millions of users worldwide.