Inside the Rise of Enterprise AI Model-Switching

Perplexity is the only enterprise AI platform that provides access to models across multiple providers in one interface. Meanwhile, 92% of Fortune 500 companies, including six of the Magnificent 7, use Perplexity Enterprise to get answers and make decisions. This gives us a unique position to see how companies actually use different AI models for different tasks and teams. 

The 2025 data reveals a clear pattern: aside from engineering teams, no model achieved greater than 17% share of overall usage, and as AI models and users became more sophisticated, usage between models became increasingly fragmented. In January 2025, two top models accounted for more than 90% of enterprise usage. By December, four models had more than 10% share each, and even the leading model captured only about 23% of queries.

The rise of the multi-model workforce

Enterprise usage data shows employees are now treating models as a menu of options, not a single default tool. Among enterprise seats, 12.5% qualify as active power users, engaging at least 12 out of every 28 days. These power users drive model exploration: 40% actively use 6+ models, compared to 20% of regular users. The pattern scales at the organizational level too. The top 50 enterprise accounts use 30 models on average versus seven for typical accounts. 

How enterprise teams match models to tasks

Leading AI companies launched 46 new models in 2025, and Perplexity offered all of them within 24 hours of release. This aggregator approach solved a major workflow problem: employees can switch between models without juggling logins or managing separate contracts.

Across enterprise customers, 43.6% of organizations used more than one model at some point during 2025, reflecting a decisive shift away from single-model standardization and toward multi-model use. The 9.1% of enterprise users who used multiple models on a single day point to power-user behavior of routing different tasks to different models. Among users who choose specific models, 53% switched models within a single workday at least once in 2025, underscoring how routine model-hopping has become for this group.

Model specialization by use case

Throughout 2025, enterprise users selected a Claude model for 38% of programming queries. At the org level, 40% of enterprises default to Claude for programming, while 22% prefer GPT.

But programming is the exception. For every other function, the lead is less clear. In December 2025, preferences varied by task type:

• Visual Arts: Gemini Flash (40%)

• Financial Analysis: Gemini 3.0 Pro Thinking (31%)

• Debugging: Claude Sonnet 4.5 (30%)

• Software Development: Claude Sonnet 4.5 (29%)

• Legal/Court Cases: Claude Thinking models (23%)

• Medical Research: GPT-5.1 Thinking (13%)

As new models launch, these preferences are likely to shift. What leads today may not lead next quarter.

The model landscape continues to evolve

At the beginning of 2025, Claude Sonnet 4 and GPT-4o together accounted for 91.5% of enterprise queries (47.5% and 44% respectively). By late 2025, the top models split more evenly across providers: 23.3% Gemini 3.0 Pro Thinking, 20.6% Claude Sonnet 4.5, 10.7% Claude Sonnet 4.5 Thinking, and 7.9% GPT-5.

New models typically spike above 50% of enterprise usage for a few days following their release, as users experiment with new capabilities (GPT-4.1 in late April, GPT-5 in early August, GPT-5 Chat in late October). That quickly tapers to 35% at most by the following week. Models go in and out of favor as new releases arrive and teams refine which model fits which job.

Why flexibility beats picking a single winner

Claude dominated for engineers, but that was the exception. Across every other function, teams split across different models depending on the task. When the landscape shifts this quickly, access to every option matters more than choosing today's best one.