Seedance 2.5 is about to launch—Will long-duration native features and local editing make up for commercial shortcomings?

```

On June 23, Volcano Engine held the 2026 Summer FORCE Primal Power Conference in Beijing, where it released the Doubao Large Model 2.1, as well as several new models in video, image, audio, and other fields, and upgraded its Agent-oriented cloud service system.

According to Volcano Engine, as of June this year, the daily average token calls for the Doubao large model have surpassed 180 trillion, a more than tenfold increase over the past year. IDC data shows that in China's public cloud MaaS service market, Volcano Engine ranks first with a market share of 49.5%.

As the core model released this time, Doubao 2.1 Pro has achieved a leap in capabilities in three main directions: Coding, Agent, and VLM. Volcano Engine claims that its performance surpasses Claude Opus 4.6 in several evaluations, officially crossing the production-level "tipping point."

Evaluation results show that Doubao 2.1 Pro entered the first tier in code evaluations such as Terminal Bench 2.1, SWE-Pro, and SciCode; in Agent and multi-modal areas, the model also ranked among the top globally in evaluations such as OSWorld, MobileWorld, and MMMU-Pro.

At the conference, Volcano Engine president Tan Dai demonstrated Doubao 2.1 Pro’s capabilities in real engineering tasks. In a chip design RTL test, Doubao 2.1 Pro ran continuously for nearly 18 hours, went through nine iterations, and completed the full engineering process including simulation, testing, and comprehensive inspection.

In terms of pricing, Doubao 2.1 Pro charges 6 yuan per million tokens for input, 30 yuan for output, and 1.2 yuan for cache hit.

Volcano Engine stated that its total usage cost is nearly 80% lower than that of Claude Opus 4.6. For high-frequency call scenarios, Doubao 2.1 Turbo’s price is further reduced to half that of 2.1 Pro.

However, the most anticipated highlight of this conference was the debut of the latest generation video generation model Seedance 2.5, which is expected to launch officially in July.

This is another major upgrade to the product by ByteDance after Seedance 2.0 became a significant player in the domestic AI video generation field.

According to the introduction, Seedance 2.5 has achieved three main improvements in video generation capabilities: First, it supports direct output of native video segments up to 30 seconds in length; second, it supports joint generation of up to 50 full-modality assets; third, it supports local editing that maintains visual consistency.

These upgrades are aimed at the usability of video generation models in real production scenarios.

Improvement in the ability to output 30-second native video segments means the model’s capabilities of temporal consistency and narrative continuity are further enhanced;

The improvement in local editing capability solves another key pain point in the implementation of AI video in practice. For enterprises and professional creators, post-generation editing means users can modify characters, backgrounds, objects, or local details without having to regenerate the entire video segment, thus lowering the cost of trial and error and improving production efficiency.

Tan Dai believes that video generation is one of the paths to a world model and has huge application potential in the real economy. At present, Seedance has already been applied in embodied intelligence, industrial manufacturing, intelligent driving, and other fields, providing new tools for business needs such as data synthesis, scenario simulation, and process demonstration.

As video models begin to penetrate scenarios such as embodied intelligence, industrial manufacturing, and intelligent driving, the key focus of the next stage of competition may be on who can more stably serve real industrial processes.

Risk Warning and DisclaimerThe market involves risks, and investment should be made cautiously. This article does not constitute personal investment advice, nor does it take into account the individual investment objectives, financial situation, or needs of specific users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their particular circumstances. Any investment based on this is at your own risk. ```