Introducing DLO: Dynamic LLM Orchestration for Smarter, Cheaper, Faster AI

At intuist.ai, we started by building on OpenAI, which offered a strong foundation for general-purpose generation. As we expanded use cases, we decided to add both Gemini and Claude and began noticing subtle performance differences across tasks and LLMs. Some models were faster in certain scenarios, others were more precise. Claude consistently delivered stronger results for reasoning, math, and data-heavy prompts, while Gemini stood out in summarization and vision tasks. OpenAI continued to be reliable and versatile, but it wasn’t always the best fit for every job. That’s when we realized: the future isn’t about picking a model. It’s about choosing the right one for each request.

That’s when we decided to build DLO: Dynamic LLM Orchestration.

What is DLO?

DLO is the orchestration engine that powers every agent you build on intuist.ai. Behind the scenes, it automatically picks the best large language model - OpenAI, Claude, or Gemini - for the task at hand. You don’t need to set any preferences, toggle between models, or even know which one is being used. It just works.

You ask a question. Your agent responds. DLO handles the rest.

Why Does DLO Matter?

Not all models are built the same. Each one has strengths and weaknesses:

• Claude 3 Opus consistently ranks highest on math, coding, and structured reasoning tasks.
• Gemini 1.5 Pro is optimized for multimodal input and excels at summarizing large documents.
• OpenAI GPT-4 remains highly capable for general writing, conversation, and creative ideation, but can be slower or unreliable under load.

The challenge isn’t picking the best model. It’s picking the best model for each request.

That’s what DLO solves.

Performance and Cost, Optimized

DLO routes every call to the model that offers the best combination of accuracy, speed, and cost. You don’t need to pay premium rates for simple queries, and you don’t need to worry about tuning temperature settings or managing API keys across providers.

This dynamic routing reduces costs while maintaining high-quality outputs. Our benchmarks show up to 30% savings in usage fees when compared to single-provider solutions, with better output quality for complex workflows.

Reliability Built In

Language models can go down. API endpoints fail. Latency spikes. Rather than exposing users to these issues, DLO provides automated fallbacks.

If one provider is slow or offline, DLO silently retries the same request with another model. For example, if Gemini is overloaded, DLO shifts the task to Claude. If Claude fails, it tries OpenAI. All of this happens in milliseconds- without interruption to the user experience.

Why We Built This

Originally, Intuist AI supported only OpenAI. It was fast to start, but we soon ran into model outages and quality inconsistencies. To improve stability, we added support for Claude and Gemini. That’s when the real insight hit: some models are simply better at certain things.

What started as a backup strategy became a new foundation. Instead of asking users to choose a model, we decided to choose for them: intelligently, automatically, and invisibly. That’s the heart of DLO.

What Does This Mean for You?

Honestly, not much, and that’s the point. You shouldn’t have to think about which model is being used. You’ll just notice faster responses, better answers, and fewer failures. The complexity is ours. The experience is yours.

Want to learn more? Contact the team and we'll happy to share a demo or discuss how DLO applies to your needs.