What kind of providers of AI models are there?
Variety of Providers and Models in nele.ai
nele.ai is a platform that provides a wide range of generative language models from various providers. Currently available partners include renowned companies such as OpenAI, Microsoft Azure, Google, Anthropic, AWS Bedrock, and Mistral. Furthermore, new models are continuously integrated to always offer the latest developments in artificial intelligence.
The currently available models include the following:
- Advanced models from the GPT-5 family, such as GPT-5.5, GPT-5.4, and their variants
- The proven GPT-4.1 generation with GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano
- Google Gemini 3.5 Flash, as well as Gemini 2.5 Pro and Flash
- Claude 4.8 Opus, Claude 4.6 Sonnet, and Claude 4.5 Haiku from Anthropic
- Mistral Large 3, Mistral Medium 3.5, and Mistral Small 4 for specialized applications
-
Through these partnerships, nele.ai consistently provides the latest AI models from these providers, ensuring compliance with high security and data protection standards.
European Server Locations: An Alternative to the US
It is particularly important that powerful AI models are also available on European servers. This enables organizations that must keep their data within Europe to utilize these models without restriction. For this reason, nele.ai provides nearly all AI models on European servers.
Available models on European servers include:
- GPT-5.5, GPT-5.4, GPT-5.1, and the entire GPT-4.1 series via Azure infrastructure
- Google Gemini 3.5 Flash, as well as Gemini 2.5 Pro and Flash
- Claude 4.6 Sonnet and Claude 4.5 Haiku via Google and AWS Bedrock
- Mistral Large 3, Mistral Medium 3.5, and Mistral Small 4
Enhanced AI Model Capabilities
The available AI models offer various specialized functionalities:
Reasoning Capabilities
Modern models such as GPT-5.5, GPT-5.4, GPT-5.1, Gemini 3.5 Flash, Claude 4.8 Opus, Claude 4.6 Sonnet and their variants possess advanced reasoning capabilities, which enable complex, logical conclusions and multi-step problem-solving.
In nele.ai, for supported models, the so-called Thinking Mode can be configured. This determines how thoroughly the model processes a request before responding. The following levels are available:
- Standard Configuration – provider-predefined configuration
- Brief Thinking – few simple steps for a quick answer
- Moderate Thinking – several steps, better reasoned, somewhat more detailed
- Thorough Thinking – many steps, checks assumptions, very detailed
- Very Thorough Thinking – many steps, checks assumptions, very detailed
It should be noted that not all models with reasoning functionality support the full spectrum of available thinking mode levels.
Vision Functionality
Most current models support vision capabilities for analyzing and processing image content, including all models from the GPT-4.1 and GPT-5 families, Gemini 2.5 and Gemini 3.5 Flash, as well as all Claude models.
Multimodal Applications
In addition to text models, specialized models are available:
- Image Generation: GPT Image 2 (OpenAI) as well as Gemini 2.5 Flash Image (Google)
- Audio Processing and Transcription: Azure Speech - Fast Transcription for Speech Recognition and Transcription
- Audio processing and transcription: Azure Whisper and Azure Speech - Fast Transcription for speech recognition and transcriptionAudio processing: Whisper for voice recognition and transcription
AI Models and Token Sizes in Chat Context
The AI models offered vary in their token sizes within their chat context (see also our blog post on the difference between Knowledge Base (RAG) vs. Chat Context), with extended context sizes ranging from 128k to over 1 million tokens. It's important to understand what a token is: the smallest units that make up AI models. These units can be letters, syllables, abbreviations, or entire words. Tokens are comparable to puzzle pieces that are assembled to form responses.
Context sizes vary by model:
- Medium contexts: 128k–256k tokens (Mistral Medium 3.5, as well as Mistral Large 3 and Small 4)
- Extended contexts: 200k–256k tokens (Claude 4.6 Sonnet, Claude 4.5 Haiku)
- Large contexts: 400k tokens (GPT-5, GPT-5-mini, GPT-5-nano, and GPT-5.1)
- Maximum contexts: over 1 million tokens (GPT-5.5, GPT-5.4, the GPT-4.1 series, as well as Gemini 2.5 and Gemini 3.5 Flash)
The chat context limits the number of tokens that can be processed in a chat. As a rule of thumb, 750 English words correspond to approximately 1,000 tokens, while in German, 1,000 tokens correspond to about 350 words.
Billing through our flexible and transparent pricing model
The costs for AI models at nele.ai vary depending on the model used and the number of tokens or words. For language models, billing is per token, while for image models, the price depends, for example, on the desired image resolution. nele.ai has introduced a flexible and transparent credit-based pricing model.
A particular advantage of nele.ai is its usage-based billing instead of fixed monthly fees per member. This allows every employee in an organization to gain access without incurring flat-rate costs. Recognizing that the demand for generative AI varies among employees, this model ensures fair and appropriate costs.
A key factor in this model is the AI volume consumption factor, which describes the ratio of costs to credit consumption. These factors vary depending on the model and its performance, with newer and more powerful models typically having higher factors.
This structure enables effective cost optimization and improved resource utilization.
The administration interface (manage.nele.ai) of nele.ai also simplifies the management of available AI models and their associated costs. Administrators can define which AI models are accessible to their team and restrict individual members' AI usage volume.