Major AI market share shift revealed: DALL-E plummets 80% as Black Forest Labs dominates 2025 data

New information is dramatically emerging You have a market share It turns in 2025 with quick changes used by businesses and consumers from artificial intelligence. PoeA platform hosting over 100 AI model, a a comprehensive report Today, it provides an unprecedented appearance of the real world used between text, image and video generation technologies.

Poe AnalysisBased on the interactions of millions of users in the last year, technical decision makers usually offer technical decision-makers in the field of competition. “As the AI ​​models continue to progress, they will be the center of people who acquire knowledge, how their complex tasks acquired and how their daily works.

The findings emphasize that it is an important market split between all aspects of the AI. As players like Open and Anthropical Protect dominant positions in text production, for example, newer applicants Depth In the text and Black forest laboratories Despite the mass investments flowing towards industrial leaders in description, the dynamic ecosystem, despite the dynamic ecosystem, quickly captures the proposal of a meaningful ecosystem.

Behold, Poe has five surprisingly five ways from the analysis of the AI ​​ecosystem in early 2025:

1. Google shows unequal performance between AI modals

Different AI modals, Google’s performance reveals the difficulty of achieving modal leadership. Her Gemini family Text models “Share increased message by October 2024, 2024”, but it has decreased despite an important investment and technical capacity. “

This contradicts sharply with the performance of Google in other categories. In the description generation, Google’s Image3 The family, while in the video generation, provided its effective 30% market share I-2 Model Rapidly captured 40% of the messages.

This mixed performance shows that only technical excellence does not guarantee the market management. For the entity, this determines the importance of evaluating the basis of modality and modality instead of determining the leadership of AI opportunities in an area.

2. Video generation is experiencing competition at high speed

Video generation of video, which is the newest border in the generative AI, is already a strong competition and rapidly rapidly changing leader positions. According to the report, the video generation category has expanded to more than eight providers, which now offers various options for subscribers, when only in the late 2024. “

RunwayEarly a pioneer, “Although only one API model, the video has taken a strong position with 30-50% of the gene messages. However, Google’s access immediately broke the status quo:” Google’s I see-2From the date of the last date of the Poe, the total video has captured about 40% of the gene messages in several weeks. “

Chinese advanced models make up about 15% of video generation messages. Models like “Kling-Pro-v1.5, Hailuo-ai, HunyuAnvideo and WAN-2.1 Continue to push the boundaries, the time and cost of results, “the international competition remains a significant factor in the field of driving in the field of innovation, despite geopolitical tensions of international competition.

3. Image production has a radical transformation

The description area is perhaps the most dramatic market change in the generative AI, and the established players lose their place for newcomers. “First Mover Image Gen Models -Electronics from 3 and varied Style The versions were pioneers in the space, but the official image has increased by the number of relative use shares as the number of gene models increases between 3 and ~ 25.

Black forest laboratories The surprise appeared as a leader: “Black forest laboratories Flux family Description Generation models exploded on the stage in mid-2024 and maintained their dominant position as a clear Frontunner near 40% of the press. “This is a remarkable achievement for relative newcomers against certain competitors with extensive resources.

Google’s strategic investment in the offspring is also fruit, “Google’s Image3 The family has been constantly growing since the last of 2024, almost 30% of the use of. “These positions are a strong second place controller, despite the next market entry.

Poe’s data reveals a trend for AI companies that invests a lot to protect the old models.

This sample reveals the companies with rapid leaving users GPT-4 for GPT-4O and Claude-3 for Clod 3.5. The effect is clear: protect the return and support the inheritance models, users may have returns as users move to the latest victims.

Companies, more frequent updates with various opportunities and price points, more frequent updates are more frequent updates, they may need to reconsider the strategies of products throughout the life.

5. Text AI Duopary faces new challenges

Open and Anthropical Prior to text production, but face the growing pressure of newer applicants. According to Poe, the use of text between Openai and anthropical models, demonstrating the growing competition in the highly expressive text model, since the beginning of 2024, the two companies order about 85% of the interactions on the platform.

The rapid rise of the anthropy shows that parity and abilities with Openai can quickly translate to market sharing shifts in an area with strong network effects and first movement advantages.

It is more interesting, Deepseek’s legitimate is that it appears as a third rival. The report notes that “DeepSeek-r1 and -V3 In December 2024, without using any use, went to get 7% of the higher-term messages from the previous open source model family Llama And Mistral“This dramatic rise shows that access barriers for new text can be lower than previously expected.

The way to the upcoming: A liquid view for AI decision makers

This Poe Report The technical excellence lightens a generative AI market characterized by a rapid evolution that does not guarantee the ongoing market leadership alone. An entity decision-makers are an increasingly complex seller’s view that today’s dominant player can be tomorrow-ran.

The clean thing is that the user options can change user options with new model releases, organizations should set up flexible AI piles, which can adapt to changes in single vendor solutions. AI adoption capability – text, imaging and videos make the enterprise strategy even more difficult with various leaders.

As a result, as PoE notes, “We hope that these findings offer an idea of ​​changing dynamics of the AI ​​model landscape.” The message for these fluid ecosystems navigating enterprises is unavailable: AI revolution rewards those who bet on those who bet on their technology leaders in yesterday, and reward those who are comfortable.