AI Model Showdown: Top Choices For Text, Image, & Video Generation via @sejournal, @MattGSouthern

With truthful galore AI models disposable today, it’s pugnacious to determine wherever to begin. A caller survey from Quora’s Poe provides guidance for those unsure astir which models to choose.

The study analyzes millions of interactions to item the astir fashionable tools for generating text, images, and videos.

With astir each tech institution offering an AI solution, it’s casual to get overwhelmed by choices. Poe’s information clarifies which models are trusted and wide used.

Whether you’re caller to AI oregon experienced, this study shows trends that tin assistance you find the champion models. Remember that this information represents Poe subscribers and whitethorn not bespeak the broader AI community.

Text Generation Trends

A Two-Way Race

The survey shows that among Poe subscribers, Anthropopic models are rapidly becoming arsenic fashionable arsenic OpenAI, particularly aft the merchandise of Claude 3.5 Sonnet. The usage of substance models from some providers is present astir evenly split.

Rapid Adoption of New Releases

Poe users often power to the latest models, adjacent if loyal to a circumstantial brand. For example, radical rapidly determination from OpenAI’s GPT-4 to GPT-4o oregon from Claude 3 to Claude 3.5.

Emerging Players

DeepSeek’s R1 and V3 person captured astir 7% of the messages connected Poe. Google’s Gemini household has seen a flimsy diminution successful usage among Poe subscribers but remains a cardinal player.

Image Generation Trends

Market Share of Early Movers

DALL-E-3 and StableDiffusion were erstwhile leaders successful representation generation, but their shares person dropped by astir 80%. This diminution occurred arsenic the fig of representation procreation models accrued from 3 to astir 25.

Leading Models

The FLUX household from BlackForestLabs is present the starring representation model, holding a astir 40% share, portion Google’s Imagen3 household has astir a 30% share.

Smaller Models

Smaller representation providers similar Playground and Ideogram update their services frequently, which helps them support a loyal idiosyncratic base. However, they lone relationship for astir 10% of Poe’s representation procreation usage.

Video Generation Trends

An Emerging Industry

Video procreation was astir nonexistent connected Poe until precocious 2024, but it has rapidly grown successful popularity. Now, astatine slightest 8 providers connection this ability.

Runway: Most Used Model

Runway’s azygous video exemplary handles 30–50% of video procreation requests. Although its usage is little than it utilized to be, galore radical inactive take this brand.

New Player: Veo-2

Since launching connected Poe, Google’s Veo-2 has gained astir 40% of the market, showing however rapidly lawsuit preferences tin change. Other caller models, specified arsenic Kling-Pro v1.5, Hailuo-AI, HunyuanVideo, and Wan-2.1, person captured astir 15% of the market.

Key Takeaway & Looking Ahead

The information shows a wide signifier of newer models replacing older ones successful idiosyncratic preference. If you privation the champion performance, usage the latest mentation alternatively than instrumentality with acquainted but outdated models.

Whether these usage patterns volition clasp dependable oregon proceed to displacement remains to beryllium seen. At immoderate point, outgo volition beryllium a obstruction to adoption, arsenic caller models thin to get much costly with each release.

In aboriginal reports, Poe plans to stock insights connected however antithetic models acceptable assorted tasks and terms points.

Featured Image: stokkete/Shutterstock