Volc Engine Launches the Industry's First Agent Package: Integrating Multimodal Models and Network Tools

On May 11, ByteDance's Volcanic Engine officially launched the industry's first "Agent Package" - Agent Plan, marking the acceleration of AI applications from single programming assistance to general agents (General Agents). This plan breaks through the traditional single-dimensional model subscription, deeply integrating ByteDance's self-developed full-modal SOTA models such as Doubao-Seed, Doubao-Seedance, and Doubao-Seedream, as well as mainstream third-party models like GLM-5.1 and Kimi-K2.6.

In terms of technical architecture, the Agent Plan innovatively introduces a "Model and Harness" dual-driven mode. In addition to providing multi-modal model capabilities, it also integrates Harness tools such as online search and Vision Embedding, significantly enhancing the real-time information acquisition and long-term memory capabilities of the Agent. To address developers' demand for billing transparency, Volcanic Engine introduced a unified resource measurement unit "AFP (Agent Fuel Value)". Currently, the plan offers four tiered subscription combinations, ranging from the basic version to the Max version. Users of the Medium and above versions can experience the cloud-based tool ArkClaw with deep optimization for free.

The launch of the Agent Plan not only lowers the development threshold for complex Agent applications by offering one-stop resource configuration, but also reflects that the focus of competition among large models is shifting from "basic model parameters" to "toolchain integration and ecological implementation." Through the in-depth collaboration between Model and Harness, Volcanic Engine is attempting to build a standardized AI productivity supply model, further accelerating the explosion of Agent in diverse productivity scenarios such as short video production and automated office work.

ByteDance Launches the Full-Modal Large Model Doubao-Seed-2.0-lite: AI Can Listen, Watch, and Directly Get Things Done

Volc Engine, a subsidiary of ByteDance, has released Doubao-Seed-2.0-lite, the first full-modal understanding model in the Doubao Large Model family. It achieves native unified understanding of video, images, audio, and text, breaking through the limitations of single-modal understanding. The model performs outstandingly in visual and logical reasoning capabilities, especially in complex reasoning tests in advanced disciplines such as physics and medicine, where its performance significantly surpasses existing levels, marking a key advancement in the field of multimodal interaction.

Volc Engine Launches Agentic AI Automotive Solution, Doubao Large Model Accelerates Intelligent Cockpit Ecosystem Transformation

At the 2026 Beijing Auto Show, Volc Engine launched a new generation of automotive AI solutions based on the Agentic AI architecture, integrating AI cockpit packages and the Doubao Cockpit Assistant, driving the automotive human-computer interaction from passive dialogue to the era of autonomous agents. Tao Dai, President, revealed that it has covered 100% of mainstream car manufacturers, with smart cars equipped with the Doubao large model exceeding 7 million, involving more than 50 brands and 145 models.

Volc Engine Launches Doubao Speech Recognition Model 2.0 to Improve Multilingual Recognition Accuracy

Volc Engine launches Doubao Speech Recognition Model 2.0, significantly enhancing inference capabilities and supporting multilingual and visual information recognition. The model is based on a 2 billion parameter audio encoder, optimized for complex scenarios, improving the accuracy of recognizing proper nouns, names, places, and polyphones.

Volc Engine Ranks First in China and Fifth Globally in Gartner's Report on On-Premise Capabilities

Gartner released its first Magic Quadrant for AI Development Platforms, with Volc Engine ranked as the top challenger. It has the fifth strongest on-premise capabilities globally and first in China. Its strengths lie in a complete loop of models, tools, computing power, and scenarios, enabling leading customers in industries such as consumer goods and finance to quickly build multimodal applications. By the first half of 2025, Volc Engine's market share in large model services on public clouds in China reached 49.2%, capturing nearly half of the Chinese market.

Volc Engine Launches Doubao Large Model 1.6 Upgrade, Daily Token Requests Exceed 30 Trillion!

Volc Engine upgrades the Doubao Large Model matrix and launches the 'Intelligent Model Routing' service to help enterprises move toward the Agentic AI era. Key upgrades include the Doubao Large Model 1.6 enhancing reasoning capabilities, supporting four different thinking lengths, and launching three new models: 1.6lite, Speech Synthesis 2.0, and Voice Cloning 2.0, further enriching the product ecosystem.