Figure Hark Handoff Is The Best Internet-Use Model Yet

Figure has introduced Hark Handoff, a new model designed for internet use rather than software development. The company says independent verification has ranked it as the best internet-use model available, ahead of both ChatGPT 5.4 and Claude Opus 4.8, two of the leading general-purpose AI systems on the market. Unlike many AI labs that focus heavily on coding benchmarks as a proxy for general capability, Figure has taken a different approach with Hark Handoff. The model is built to handle practical, everyday web tasks: ordering food, booking flights, shopping online and navigating websites more broadly. This suggests Figure is positioning the model as an agent for consumer-facing tasks rather than a developer tool or a general chatbot replacement. The claim of independent verification is notable, as it implies Figure has submitted Hark Handoff to third-party benchmarking rather than relying solely on internal testing. If accurate, beating both ChatGPT 5.4 and Claude Opus 4.8 on internet-use tasks would mark a significant result for a company better known for its work in robotics than in software agents. Figure has built its reputation primarily around humanoid robots, making this move into browser-based AI agents a notable expansion of its product line. The focus on internet-use rather than coding reflects a broader shift some companies are making toward agentic AI that can complete real-world digital tasks on behalf of users. Rather than optimising for benchmarks that measure programming ability, Figure appears to be betting that the more immediate commercial opportunity lies in automating routine online activities that consumers and businesses handle daily, such as travel bookings and online purchases.