Cognis Ai (Beta) Update Note: New LLMs and Custom Configurations are Now Live on Cognis Ai

This week, we’ve added new LLMs, including GPT 5.1, GPT 5.2, Gemini 3 Pro, and Gemini 3 Flash. Now you can also use custom configurations to tune your LLMs' responses in terms of reasoning, summary, temperature, and verbosity.

Across recent interactions we discovered a need for optimizing model and token usage across LLMs. While newer models like Gemini 3 Pro and GPT 5.2 outperform predecessors on benchmarks, the ability to expend all reasoning capabilities of such powerful models, is limited.

When not using custom configuration it leads to:

This difference is noticeable when executing heavier tasks using LLMs like Claude, Gemini, or GPT. Using custom configuration allows you to bypass these bottlenecks and unlock cost savings of up to 40%, depending on usage type and limits.

Custom configuration allows you to edit a model’s internal tuning across six key areas:

1. Response Mode: Allows you to control how an LLM shapes its responses. You can change it across the following levels to achieve higher levels of precision or expressiveness:

2. Reasoning: Controls the level of internal reasoning effort expended by a model before answering any query. The two levels of reasoning, as follows, are best suited to handle particular scenarios:

3. Summary: This feature tracks the internal reasoning of models. You can switch it on or off depending on the level of tasks that you assign to an LLM agent. We recommend keeping it turned on for complex multi-step tasks, allowing you to interrupt the model if it takes wrong steps.

4. Temperature: The temperature of a model varies between 0.0 and 2, as shown in the image below, with 0.0 being the most deterministic response level and 2 being the most creative and diverse response level.

You can adjust the model temperature depending on the kind of task you’re undertaking. For instance, creative tasks like writing blogs tend to work better with temperatures above 1, while technical writing tasks work better with temperatures below 1.

5. Verbosity: Controls the level of detail the model includes in its responses. It has three levels depending on the length of output that you want:

6. Thinking Budget*: This feature is only available for Gemini 2.5 Pro and governs how much internal planning the LLM performs before it begins generating a response. 128 is the lowest thinking budget, and 32,768 is the highest thinking budget available.

The new LLM models added to Cognis Ai include:

With the latest addition of models and the introduction of custom configuration, you can now optimize your token usage to spend less on task execution via LLMs on Cognis Ai.

While custom configuration amongst models is pretty much standardized, they can vary by model generation. Below is a table for reference on the capabilities of each model.

Getting Started

You can Get Started for free today with Cognis Ai’s Community Plan. Please visit Cognis Ai and click on Get Started to begin your journey.

Once you enter Cognis, you can use the quick setup to activate the popular LLMs in your account. Further, you can also configure more LLMs from the ‘Explore LLMs’ tab on the bottom left.

Our Commitment to Human-AI Collaboration

With Cognis Ai, we focus on developing a complementary AI system that helps you derive tangible results from AI. Our ethos while building Cognis is to advance the state of Human-AI collaboration to help people execute faster, remove redundancies, and focus on meaningful work that moves the needle forward. Our future roadmap intends to add features, integrations, and tools that help reduce the noise surrounding AI and focus on outcomes.

For updates, follow us on LinkedIn.