A new report from Poe provides insights into AI model usage trends from January to May 2025, showing significant shifts in user preferences across text, reasoning, image, video, and audio generation models. The report draws on aggregated usage data from Poe’s platform, which offers users provider-agnostic access to various AI models.
Assistants: According to the findings, OpenAI’s GPT-4.1 family and Google’s Gemini 2.5 Pro quickly gained market share following their launches, reaching approximately 10% and 5% respectively within weeks. Meanwhile, Anthropic’s Claude models experienced a notable decline of around 10% during the same period.
Reasoning AI: The report highlights a growing interest in reasoning models, with their share of text messages on Poe increasing from 2% to 10% during the reporting period. Google’s Gemini 2.5 Pro made particularly impressive gains in this category, capturing about 30% of reasoning message share within six weeks of launch.
OpenAI continues to dominate the reasoning model space, releasing multiple capable models in the first four months of 2025, including o1-pro, o3-mini, o3-mini-high, o3, and o4-mini. Users appear to quickly adopt newer models as they become available.
Read more: Understanding Reasoning AI: What it is, how to use it
Image generation: In image generation, competition is intensifying. OpenAI’s GPT Image Generation (GPT-Image-1) quickly gained 17% market share within two weeks of its API launch in late April. Google’s Imagen3 family saw steady growth from 10% to 30%, while Black Forest Labs’ FLUX models maintained their leadership position with about 35% share, though this represented a decline from 45% earlier in the reporting period.
Video generation: The video generation landscape witnessed a significant disruption with the emergence of Kling 2.0 from Chinese lab Kuaishou, which captured 21% of all video generation on Poe within just three weeks of its release. Google’s Veo 2 maintained a strong position with approximately 20% share, while incumbent leader Runway saw its usage share decline by about 40% to 20%.
Audio generation: For audio generation, ElevenLabs remained dominant with approximately 80% of all text-to-speech requests, though new competitors like Cartesia, Unreal Speech, PlayAI, and Orpheus are beginning to emerge with unique offerings.
The report suggests that reasoning capabilities will continue to be a major competitive driver among leading AI labs, while multimedia generation is becoming increasingly competitive.