The artificial intelligence market is undergoing an unprecedented change. According to the latest comprehensive report released by the Poe platform, the competitive landscape in the AI field has changed significantly in early 2025, and emerging companies are rapidly eroding the market share of veteran giants. This trend not only reflects the rapid iteration of technology, but also reveals the strong market demand for innovation and diversity.
As a platform with more than 100 AI models, Poe provides us with a unique perspective on the usage patterns of text, image and video generation technology based on interactive data from millions of users over the past year. These commonly strictly protected usage data provide valuable market insights for technology decision makers and help them better understand user needs and technology trends.

While established companies such as OpenAI and Anthropic still dominate the text generation field, new entrants such as DeepSeek and Black Forest Labs have quickly seized a large market share, showing the vitality of this ecosystem. This market dispersion and the rise of emerging forces mark an era of more diversification and fierce competition in the AI field.
There are huge differences in Google's performance in different AI models. Its Gemini series text models have been growing until October 2024, but have begun to decline since then. By comparison, Google's Imagen3 series in the field of image generation accounted for 30% of the market share, while its video generation model Veo-2 quickly gained 40% of the market. This imbalanced performance shows that technical strength does not guarantee a comprehensive market leadership, and corporate decision makers need to evaluate AI capabilities one by one, rather than assuming that leading in a certain field means overall excellence.
Video generation, as the latest frontier of generative AI, has witnessed fierce competition and rapid transformation of leadership. This category that only appeared at the end of 2024 has rapidly expanded to more than 8 providers. As an early pioneer, Runway has maintained a strong position of 30% to 50% despite only one API model. However, Google's Veo-2 has captured nearly 40% of the market in just a few weeks since its launch. The models developed in China account for about 15% of the video generated information. Models such as Kling-Pro-v1.5, Hailuo-AI, HunyuanVideo and Wan-2.1 have continuously broken through their limits in terms of capabilities, reasoning time and cost, indicating that international competition is still an important driving force for innovation.
The field of image generation is probably the most significant area where the AI market changes. DALL-E-3 and various stable diffusion versions as pioneers, their relative usage share dropped by nearly 80% as the official image generation model increased from 3 to about 25. Surprisingly, Black Forest Labs' Flux series model has remained significantly ahead since its popularity in mid-2024, accounting for nearly 40% of the market. Google's Imagen3 series has grown steadily since its launch in late 2024, accounting for nearly 30% of its use share, becoming a strong second-place competitor.
Poe's data reveals a trend worth noting: With the release of more powerful models, the use of new flagship models will quickly snatch old versions. Users quickly switched from GPT-4 to GPT-4o and from Claude-3 to Claude3.5. This phenomenon means: maintaining support for the old model may lead to diminishing returns. Businesses may need to rethink product lifecycle strategies, focus resources on fewer models and make more frequent updates.
OpenAI and Anthropic maintain their dominance in the field of text generation, accounting for about 85% of text interactions on the platform. Since the release of Claude3.5Sonnet in June 2024, the two companies have almost equal text usage, indicating increasingly fierce competition. What's even more striking is that DeepSeek becomes the legal third contender. DeepSeek-R1 and V3 gained 7% market share from unmanned use to peak in December 2024, significantly higher than any previous open source model. This sharp rise suggests that the entry barriers for new text AI providers may be lower than expected.
Poe's report clearly demonstrates the characteristics of the rapid development of the AI market: technological advantages alone cannot guarantee a sustained market leadership. Business decision makers face an increasingly complex supplier landscape, and today's leaders may become tomorrow's laggards. User preferences may change drastically with new models being released, suggesting organizations should build flexible AI stacks to adapt to changing capabilities rather than being limited to single-vendor solutions. AI has emerged with different leaders in text, image and video, further complicating corporate strategies.
As Poe points out in his conclusion: “We hope these findings give people a glimpse into the dynamics of the AI model landscape.” For businesses navigating this mobile ecosystem, the message is clear: the AI revolution is growing at an astonishing pace, rewarding companies that remain flexible while punishing organizations that rely too much on yesterday’s technology leaders.