Become a member of our daily and weekly newsletters for the latest updates and exclusive content about leading AI coverage. Leather
Pussy‘S newest usage report to show Openi And Google strengthening their positions in important AI categories while Anthropic Loss of land and specialized reasoning options arise as a crucial competitive battlefield.
According to Data released today by PoeA platform that offers access to more than 100 AI models made significant shifts in the field of market share between all important AI categories between January and May 2025. The data from POE subscribers, offer rare visibility in actual user preferences outside the industrial benchmarks.
“As a universal gateway up to 100+ AI models, Poe has a unique image of usage trends in the ecosystem,” said Nick Huber, Po’s Ai Ecosystem Lead, in an exclusive interview with Venturebeat. “The most surprising things that are happening now are rapid innovation (3x the number of releases Jan-May 2025 versus the same period in 2024), an increasingly diverse more competitive landscape and reasoning models are the clear success story from the beginning of 2025.”
GPT-4O maintains dominance, while new models quickly catch market share
In core text generation, OpenAi’s GPT-4O held his commander with 35.8% of the message share, while the company is newer GPT-4.1 Family Fast 9.4% of use recorded within a few weeks of launch. Google’s Gemini 2.5 Pro Similarly, about 5% reported message share shortly after the introduction.
These profits were largely at the expense of anthropic Claude modelswho saw an absolute decrease in 10% share during the reporting period. The report notes that Claude 3.7 Sonnet has now replaced the earlier substantial Claude 3.5 Sonnet With the preference of users, although the latter still maintains a remarkable use of 12%.
DeepWho experienced viral growth earlier this year, has seen his momentum slowly, because competitors have released their own affordable, extensive reasoning models. Deepseek R1The message share of the message fell from a peak of 7% in mid -February to 3% at the end of April.
Complex problem-solving possibilities become an important differentiator on the AI market
Perhaps the most important trend that is identified in the report is the dramatic growth in specialized reasoning models, which have been extended from approximately 2% to 10% of all SMS reports that have been sent on Poe since the beginning of 2025.
“Reasoning models, even in the early days, have shown a remarkable capacity to be able to handle complex tasks with increased precision,” Huber told Venturebeat. “Early adopters find clear value in this and are willing to take on the considerations in cost and processing time for better results.”
In this fast -growing segment, Gemini 2.5 Pro Has quickly established itself as a leader and has recorded approximately 31% of the reasoning model use within only six weeks of launch. It now leads the category, prior to Claude’s reasoning -specialized models.
Openi continues to innovate quickly in this space and releases several reasoning models (O1-PRO” O3-Mini” O3-mini-high” O3And o4-mini) Just in the first four months of 2025. The report indicates that POE users quickly accept the latest offer from OpenAi, which switch from older models such as O1 to newer alternatives such as O3.
The report also noticed the rise of hybrid reasoning models, such as Gemini 2.5 Flash Preview And Qwen 3That can dynamically adjust their level of reasoning within conversations. However, these models currently represent only about 1% of the reasoning model use.
Industrial analysts suggest that this shift to specialized reasoning options indicates a ripening AI market where raw text generation is being constructed, forcing providers to distinguish them by higher value possibilities that premium prices can order.
Google’s imagen 3 -Provision -challenges based players in Visual Ai Arena
The market for image generation seems increasingly competitive, with that of Google Image 3 Family is growing steadily from approximately 10% to 30% share in 2025, now competing category leader Black Forest Labs’ Flux family of models, which together had around 35% share from the end of April.
OpenAi’s GPT-IMAGE-1introduced in the API at the end of April, reached 17% of the use of image generation in just two weeks, with the viral adoption in the Chatgpt -App During March and early April.
The report indicates that Flux models Maintained their overall pluralism share in image generation on POE, but experienced a moderate decrease of approximately 45% to 35% during the reporting period.
This three -way competition between Google” OpeniAnd Black Forest Labs marks a significant shift from the beginning of 2024, when Midjourney And Stable diffusion Variants dominated the space. The rapid improvement in image quality, compliance with prompts and the rendering speed has transformed this category into one of the most brightly disputed AI Battlegrounds.
Enterprise acceptance of image generation has been considerably accelerated in the past six months, according to additional data from industry, with marketing departments and creative agencies that increasingly integrate these tools into their production workflows.
Chinese Upstart Kling disrupts the Video AI market and challenges the early lead of the runway
In Vide Operation, Chinese Laboratory Kuaishous Newly released Kling -Family The market quickly disrupted models and jointly recorded around 30% user share. The most striking thing is that Kling-2.0-Master reached 21% of all video generation on Poe at the end of April, just three weeks after the release.
Google’s I see 2 maintained a strong position with approximately 20% share after the launch of February, while category pioneer Run Saw the user share considerably decrease from approximately 60% to 20% during the reporting period.
The speed of Kling’s Market penetration emphasizes how quickly the competitive landscape can shift in emerging AI categories, where established players may not retain their early benefits, because newcomers repeat and improve quickly.
Video generation remains the most arithmetic intensive AI application with consumer-oriented AI, with models that require considerable processing capacity to make even short clips. This has kept the use more limited than text or image generation, but rapidly falling costs and improving quality is expected to stimulate a wider acceptance until 2025.
Early adopters of companies include advertising agencies, makers of social media and educational platforms that have started integrating AI-generated video in their content strategies despite the current limitations of technology.
Elevenlabs dominates Voice AI, while newcomers focus on specialized use cases
Elfs Continues to lead the Audio Education category and fills around 80% of the text-to-speech requests from all subscribers during the reporting period. However, the report emphasizes emerging competition from newcomers Cartesia” Unreal speech” PlayiAnd Orpheusthat offer differentiated speech options, effects and price models.
This market dominance by a single player is in stark contrast to the more fragmented competition in other AI categories. Industry experts attribute the continuous leadership of Elevenlabs to its early market input, an extensive speech library and consistent quality improvements that have maintained a technical lead over competitors.
Newer participants are successful by focusing on specific market niches. Unreal speech Got grip with podcast producers and audio book publishers by offering specialized voice actors and emotional access opportunities. In the meantime, Cartesia has focused on multilingual voices with authentic accents, the capital of interests of global companies and educational platforms.
The Audio AI market is expected to grow considerably until 2025 as text-to-speech possibilities that approach human-like quality and find applications in customer service, accessibility solutions and content creation. The relatively low calculation requirements compared to generating video flowing ensure broader implementation and experiments.
Strategic implications for companies that navigate a constantly shifting AI landscape
The dynamic nature of the AI model landscape offers both opportunities and challenges for companies that integrate these technologies.
“It can be a challenge to keep track of the newest AI and the pace only wins speed,” Huber told Venturebeat. “If you are a company that is already working on scale AI, investing in robust, provider-agent evaluation pipelines is crucial because the model that is the best this month is on the second best next month.”
This volatility in model preferences underlines the value of platforms such as POE that offer access to multiple models via a single interface, allowing users to compare outputs and adapt to the changing AI eco system.
Industrial analysts suggest that the growing importance of reasoning possibilities can indicate a shift in how companies evaluate and implement AI models, with an increasing focus on precision and reliability for complex tasks instead of just speed or cost efficiency.
As Frontier Labs continue to release more capable models at an accelerating pace, companies are confronted with difficult decisions about when they should standardize on specific platforms versus maintain flexibility. Many Enterprise AI leaders use a portfolio approach, use different models for different tasks while retaining the possibility of changing providers as the possibilities evolve.
“This will be an important space to look at, especially among border providers, because it represents the best of what AI can currently achieve,” Huber noted with regard to reasoning models.
The report indicates that multimedia possibilities are also becoming more and more competitive, which suggests that text generation, long the primary focus of AI development, perhaps makes way for a more balanced ecosystem where image, video and audio trial play equally important roles.
Companies that are successfully navigating through this complex landscape will probably be those who maintain evaluation frameworks that are aimed at specific use cases instead of boosting the latest model releases, while at the same time building technical infrastructure that makes rapid acceptance possible when meaningful improvements appear.
While AI models continue their play of musical seats on top of the ranking, one thing becomes clear: in the current market, the crown rarely remains on the same head and companies that use their future on yesterday’s AI champion, can be in line with the Ran of Tomorrow.
