What's New?

Three new models from OpenAI are now available on the Evoya platform: GPT-5.6 Sol, GPT-5.6 Terra, and GPT-5.6 Luna. The models complement the existing offering and can be used directly in your agents and workflows.

The Evoya platform offers many features – but not every user needs all of them. With the new Feature Access, you can control per role which features are visible and accessible to which users. This keeps the platform clear and tailored to individual needs.

Admins can access the page as follows: open Settings in the site menu and find it under Configurations → Feature Access.

Scheduled Tasks can now be linked to a project. The project folder serves as a central storage for all results: an agent can, for example, automatically fetch information from the internet daily, create a report, and save it directly in the project folder – or write articles and store them there. This keeps all outputs of a task collected in one place, available to the agent as context at any time, and ready for further processing. Recurring workflows thus run fully automated – and the result is not scattered somewhere, but exactly where it belongs.

Data Sources can now be linked to a folder from the File System . Instead of manually uploading documents to a data source, it's enough to place the files in the corresponding folder – the data source updates automatically. This saves time and makes document management much easier.

Projects are now available on the Evoya platform – known from tools like ChatGPT or Claude. A project gives an agent its own workspace with file access, context, and notes. This allows more complex, multi-step tasks to be handled in a structured and continuous way.

With the new Agent Builder, an AI agent can be created in just a few minutes. A guided assistant in five steps asks what the agent should do, which skills it needs, and how it should behave. No technical prior knowledge required – simply go to the Agent Overview , click "Create New Agent," and get started.

The Evoya platform has been expanded with several powerful models:

• Claude Fable and Claude Opus 4.7 from Anthropic
• Gemini 3.5 Flash from Google
• DeepSeek V4 Pro and DeepSeek V4 Flash
• GPT-5.5 from OpenAI

All models are available for selection immediately.

The new File System works like a personal cloud storage. Documents, PDFs, and other files can be organized in folders. Files generated by agents – such as Word documents, PDFs, or images – are automatically stored there. A central storage for all content – clear and always at hand.

Skills are now available on the Evoya platform. Skills let you define reusable capabilities and assign them specifically to agents – for more consistent results and more efficient workflows. The assigned skills are automatically loaded by the agent as soon as it needs them for the respective task – entirely without manual intervention.

With Outline, Exa, as well as Jira and Confluence from Atlassian, three new MCP Integrations are available. Via a simple MCP connection, these tools can be integrated directly into agents – seamlessly linking existing work environments with the Evoya platform.

Three new models for image generation are available immediately: GPT-Image 1.5, Nano Banana Pro, and Flux 2 Pro. This means even more options to choose from – from quick drafts to high-quality results.

With Qwen 3.5 and Gemma 4, two new open-source models are available, operated on Swiss infrastructure. A strong option for companies with high data protection and compliance requirements – powerful and privacy-compliant.

The new Analytics Dashboard provides a clear overview of the usage of the Evoya platform. Requests can be filtered by time period and evaluated at various levels – overall, by agent, by user, or by model. This way, you always keep track of AI usage in your company.

Emails can now be answered by full-fledged AI agents via the Email Automation – not just by simple language models. Agents have access to knowledge bases, can retrieve information, and respond in context. This makes automated email processing significantly more powerful and relevant.

Evoya agents can now truly calculate and work with data. Thanks to the new Code Interpreter, they independently perform calculations, process structured data, and create finished documents – without any manual intervention on your part.

What does this mean for your daily work? Your agent analyzes sales figures from Excel, calculates ROI and budget forecasts, prepares raw data for presentations, or automatically creates reports as Word documents. Tasks that normally cost time and manual effort now run in the background – while you focus on more important things.

How to activate the Code Interpreter: When creating or editing an agent, go to "Tools" and add the Code Interpreter.

When users chat with AI agents, chat sessions are created. These sessions are now also assigned to the workspace in which the conversation took place.

Previously, sessions were only organized by agent. From now on, the workspace is also saved with each new session. This keeps sessions of the same agent cleanly separated by workspace.

All sessions created before January 26, 2026 remain organized by agent.

Additionally, the session view allows switching between "Sessions of this agent" and "all sessions in workspace". In the workspace view, all sessions of the current workspace are displayed; when opening a session, the originally used agent is automatically selected.

Nano Banana Pro, GPT Image 1.5 & Flux 2 Pro

We have integrated the latest and most powerful AI image models into Evoya: Nano Banana Pro from Google, GPT Image 1.5 from OpenAI, and Flux 2 Pro from Black Forest Labs. With these models, you can not only generate high-quality and complex images but also edit existing images, apply various visual styles, and even create complex process diagrams with text elements – ideal for presentations, marketing materials, or internal documentation. Nano Banana Pro particularly excels at precisely displaying text within images, opening up new creative possibilities.

Forward emails to agents based on rules

Evoya has always offered email functionality that goes far beyond what providers like ChatGPT or Claude offer. Now we've added the "Process Emails" action: emails can be automatically forwarded to your agents for processing based on filter rules. They can then create summaries, update FAQs, store information in your CRM, extract and forward important points, or create tasks in project management tools. Your agents become intelligent email assistants that take routine work off your hands and ensure that important information doesn't get lost.

New top models for versatile applications

New top models are now available on the platform: Evoya Alpha (100% hosted in Switzerland – no US clouds, no data leaves the country), GPT-5.1, 5.1 Chat, and 5.1 Codex (versatile general-purpose models for virtually any task), Mistral Large 3 (European top challenger to GPT – strong in text, analysis, and data protection), as well as Claude Sonnet & Haiku 4.5 (the best reasoning models – ideal for complex tasks, analysis, and coding). Whether text, code, or analysis – with these models, you get even more out of your AI processes.

Record meetings and process them directly

Never write meeting minutes by hand again! With the new voice recording feature, you can easily record meetings or conversations – in any language, including Swiss German. The tool automatically recognizes who said what and creates a transcript with summaries, task lists, or sentiment analyses. Your agents can use the transcript to send emails or store knowledge in a knowledge base – this way, every conversation becomes real knowledge, without any extra effort.

Modify uploaded images precisely

Your Evoya agents can now not only generate images but also precisely modify uploaded or previously created images. Change clothing, alter locations, or add objects – no problem. The Flux Context model is considered the strongest image model for image modification and can turn you into a fierce pirate or heroic knight in no time.

Targeted access to AI resources

With the newly introduced Evoya Workspaces, organizations can precisely control access and use of Evoya AI resources. Departments, project teams, or school classes can create their own workspace and define members as well as resources (agents, data sources, tools). This gives each team its own, undisturbed workspace with the appropriate AI tools.

Have recurring tasks completed automatically

With the new Task Scheduler, your AI agents take over recurring tasks proactively and fully automatically. You define once what should be done and when – the rest runs by itself. Whether daily market updates, weekly reports, or reminders: simply create a task, set the schedule, and sit back. You can create tasks directly in the workspace via text or spoken language and combine different data sources, tools, and agents.

Standardized connection of software applications

The Model Context Protocol (MCP) enables diverse software systems to connect with your Evoya agents in a standardized way. We have now integrated MCP into the Evoya AI platform – making hundreds of applications available to interact with your agents. MCP is still a very young protocol (debut November 2024), but is being rapidly adopted by manufacturers and significantly expands the application possibilities of your AI agents.

Start agents via link with context information

Previously, you had to copy relevant data from your application and paste it into the chat. Now you can easily start an agent via link and provide it with context information right away. The agent can be configured to respond directly – without initial user input. These dynamic links can be very easily integrated into intranets, Excel, or Word files and significantly accelerate work processes.

Pre-configured AI agents for immediate use

Not interested in spending hours building agents? Our new library provides pre-configured AI agents for various applications – from diagram creation to build generator to researcher that scours the web with Perplexity. The best part: each agent uses exactly the language model that is best suited for the respective task. Our platform combines these strengths specifically for each use case – so you always benefit from the optimal AI agent without having to worry about technical details.

More precise results through combined search technology

The most important information is often buried somewhere in documents, presentations, or old reports – and new employees often don't even know that certain information exists. To solve this problem even more precisely, we've further developed our search: we now combine traditional full-text search with AI search. Your agents now find relevant answers even faster, more precisely, and directly from your own sources – whether PDF, Word, website, or database.

Seamless integration into your IT environment

Thanks to LDAP and Active Directory integration, you can seamlessly integrate our platform into your IT environment. You manage access and users centrally – without additional effort. This way, you maintain full control over permissions and user management while your team benefits from AI functionality.

New top models and optimized selection

New language models are now available in the Model Hub – for even more flexibility and application possibilities. Newly added: The latest Claude models from Anthropic (Opus 4 and Sonnet 4), Google Gemini 2.5 Flash for quick responses, open-source power with Qwen 3 (235B and 32B), and Perplexity Sonar for efficient web research. At the same time, we've replaced older models with faster versions. All information about available AI models can be found as always in the Model Hub directly on the Evoya platform.

Voice control for more natural interactions

You can now control your Evoya AI agents with your voice – either live or via short recordings. Start a real-time voice assistant for quick conversations or use speech-to-text: record your message, send it, and receive a written response. Perfect when you're on the go or typing is inconvenient. Have ideas how voice could help your company? Talk to us.

Save, manage, and share your best prompts

A prompt is the instruction you give to an AI – and good prompts are the key to great results. Instead of rewriting important prompts every time, you can now save, manage, and share them with your team in the new Prompt Library. Simply enter a slash ( / ) in the chat input field in the workspace to access your saved prompts. Good inputs lead to reliable results – and now your entire team can benefit from them.

New top models with up to 1 million token context

April brought powerful new additions to our LLM Hub: ChatGPT-4.1 (exclusive to our platform), o3 and o4 mini, Meta Llama 4, Google Gemini 2.5, and Grok 3 – the first xAI model on our platform. Many of them now support up to 1 million tokens of context, perfect for long documents, complex tasks, and in-depth analysis. Use the new models to achieve even better results and tackle more complex tasks.

Data protection at the highest level

Our platform already offers over 20 language models – now a special one is being added: Swiss Llama 3.3, one of the first open-source models hosted in Switzerland. This means: no data outflow abroad. We've already been storing customer data securely in Switzerland – now processing can also take place here. Llama 3.3 is among the best open-source models and is a strong option for companies that value data protection. Function calling (e.g., live internet queries) will be available soon.

New image generation at the highest level

The Evoya platform brings together the best AI models from all providers. Previously, we used DALL-E 3 from OpenAI for image generation – now you can use Flux from Black Forest Labs. The image quality is impressive. Simply set up an image generator tool, select Flux as the model, and make the new tool available to the agent of your choice – done.

Agents access your databases directly

Agents can now access your databases directly and answer questions about the data or create new records. The agent generates SQL queries appropriate to the user's question, executes them, interprets the result, and answers the question accordingly. The tool can optionally give the agent read-only access, read/write access, or even the ability to create and modify tables.

Automatic data import for even more flexibility

Evoya imports data directly from plain text, websites, or files. With the new data source type "API", data can now also be read in and vectorized via API. This allows external systems to automatically send new data to the Evoya data source – e.g., newly recorded products or CRM entries. We can also use import scripts to prepare stubborn, poorly structured content so that the data source receives optimal splits.

New models for image, text, and more

After the hot model January (keyword DeepSeek), February was quieter – nevertheless, we're providing exciting innovations with Flux, Llama 3.3 Swiss, and Claude 3.7 Sonnet. New: Flux 1.1 Pro (our new image model), Swiss Llama 3.3 Edition (completely hosted in Switzerland), Anthropic Claude 3.7 Sonnet (update of Claude 3.5 Sonnet), and Google Gemini Flash 2.0 (update of Flash 1.5). All information about available models can be found in the Model Hub directly on the Evoya platform.

The latest AI models now available

In January, we also integrated the best new AI models into the Evoya platform. New: DeepSeek R1 (an impressive reasoning model), OpenAI o1 and o1-mini (additional reasoning models), Google Gemini Pro 1.5, and Minimax 01 with a record context window of 4 million tokens. Additionally, various existing models have been updated to the latest version. All information about available models can be found in the Model Hub directly on the Evoya platform.

Pre-configured agents for faster deployment

The new library makes using AI agents even easier. We offer pre-configured agents that are optimized for industry-specific use cases and can be activated with just a few clicks.

Test and optimize agents systematically

We chat with an AI agent and are often quickly satisfied – but how do we know it couldn't be even better? With our new Agent Testing Framework, you can submit a set of test questions to multiple agents with different configurations simultaneously and compare the answers directly. Even better: if you provide a reference answer, AI can even automatically rate the agent responses. This way, you optimize your agents systematically and data-driven.

Protect sensitive data – automatically and reliably

Data security is a central concern when using AI. With Evoya AI Privacy Shield, sensitive information such as names, locations, or phone numbers are anonymized within Switzerland before being transmitted to external language models. This keeps customer data protected while you use the full power of modern AI. An example: "Max Mustermann, Zürich, 079 123 45 67" becomes "[Name], [Location], [Phone Number]".

Display visual data directly in the workspace

While generative AI models create texts, images, or videos, they often fail at meaningful graphical representations. Evoya AI now offers a solution: two comprehensive diagram libraries are available in the workspace – Mermaid and VegaLite. With VegaLite, even interactive data visualizations are possible directly in the workspace. The results are convincing and make complex information understandable at a glance.

New and updated models for even better results

In November and December, we expanded and updated the language model selection in Evoya AI. New additions: Nvidia's Nemotron (70B), Qwen 2.5 (72B) from Alibaba with support for 29 languages, and Gemini 1.5 Flash (8B) with an impressive context size of 1 million tokens – ideal for large amounts of data. Additionally, existing models like Gemma, Llama 3.2, and Llama 3.3 have been brought up to date.