DocumentationModels
Models
In anymize you choose between AI models from several makers, and this page shows for each one where it runs, what the provider does with your requests and what it can do.
5 min readUpdated
What the model choice means
With anonymization on, anymize replaces names, addresses, file numbers and other personal data with placeholders before any model sees the text. That applies to every model on this page. The model choice decides the tool, not the protection.
Smart Select picks a fitting model for each request from the models you may use. You can also pick one model in chat and stick with it.
This list is the model catalog admins see under Governance > Models. anymize's own models (Waterfall, Fountain) are not in it, because every account has them without a choice of provider. Self-connected models (BYOM) are not in it either.
Hosting: EU or international
Every model has a jurisdiction. An EU model runs in data centres in the EU. An international model runs at the maker or an operator in the USA.
Many models are available from several operators, for example Google Cloud (Vertex AI) and AWS Bedrock. For each request anymize picks the operator with the best availability, but only within the model's jurisdiction. An EU model never falls back to an operator outside the EU. If everything in the jurisdiction is down, the model is briefly unavailable instead of running somewhere else.
Some models exist in both jurisdictions. The international version then carries (USA) in its name.
All models at a glance
| Model | Maker | Hosting | Data retention at the provider | Context window | Inputs | Capabilities | Class |
|---|---|---|---|---|---|---|---|
| Qwen3 235B | Alibaba | EU | None (ZDR) | 128,000 | Text | Tools, Reasoning | Swift |
| Qwen3 32B | Alibaba | EU | None (ZDR) | not stated | not stated | not stated | Swift |
| Qwen3 Next 80B Thinking | Alibaba | EU | None (ZDR) | not stated | not stated | not stated | Swift |
| Claude 4.5 Haiku | Anthropic | EU | None (ZDR) | 200,000 | Text, Images | Tools, Reasoning | Swift |
| Claude 4.5 Opus | Anthropic | USA | None (ZDR) | 200,000 | Text, Images | Tools, Reasoning | Deep |
| Claude Fable 5 | Anthropic | USA | 30 days | 1,000,000 | Text, Images | Tools, Reasoning | Ultra |
| Claude Fable 5.1 | Anthropic | USA | 30 days | 1,000,000 | Text, Images | Tools, Reasoning | Ultra |
| Claude Fable 5.2 | Anthropic | USA | 30 days | 1,000,000 | Text, Images | Tools, Reasoning | Ultra |
| Claude Haiku 5.5 | Anthropic | EU | None (ZDR) | 1,000,000 | Text, Images | Tools, Reasoning | Swift |
| Claude Haiku 5.5 (USA) | Anthropic | USA | None (ZDR) | 1,000,000 | Text, Images | Tools, Reasoning | Swift |
| Claude Opus 4.8 | Anthropic | EU | None (ZDR) | 1,000,000 | Text, Images | Tools, Reasoning | Deep |
| Claude Opus 5 | Anthropic | USA | None (ZDR) | 1,000,000 | Text, Images | Tools, Reasoning | Deep |
| Claude Opus 5.5 | Anthropic | EU | None (ZDR) | 1,000,000 | Text, Images | Tools, Reasoning | Deep |
| Claude Opus 5.5 (USA) | Anthropic | USA | None (ZDR) | 1,000,000 | Text, Images | Tools, Reasoning | Deep |
| Claude Sonnet 4.5 | Anthropic | USA | None (ZDR) | not stated | not stated | not stated | Smart |
| Claude Sonnet 4.6 | Anthropic | EU | None (ZDR) | 1,000,000 | Text, Images | Tools, Reasoning | Smart |
| Claude Sonnet 5 | Anthropic | USA | None (ZDR) | 1,000,000 | Text, Images | Tools, Reasoning | Smart |
| Claude Sonnet 5.5 | Anthropic | EU | None (ZDR) | 1,000,000 | Text, Images | Tools, Reasoning | Smart |
| Claude Sonnet 5.5 (USA) | Anthropic | USA | None (ZDR) | 1,000,000 | Text, Images | Tools, Reasoning | Smart |
| DeepSeek V4.1 Flash | DeepSeek | EU | None (ZDR) | 1,000,000 | Text, Images | Tools, Reasoning | Swift |
| DeepSeek-V3-0324 | DeepSeek | USA | None (ZDR) | not stated | not stated | not stated | Smart |
| Gemini 2.5 Flash | USA | None (ZDR) | 1,048,576 | Text, Images | Tools | Swift | |
| Gemini 2.5 Pro | EU | None (ZDR) | 1,048,576 | Text, Images | Tools, Reasoning | Deep | |
| Gemini 3.0 Flash | USA | None (ZDR) | 1,048,576 | Text, Images | Tools, Reasoning | Smart | |
| Gemini 3.1 Flash-Lite | EU | None (ZDR) | 1,048,576 | Text, Images | Tools, Reasoning | Swift | |
| Gemini 3.5 Flash | EU | None (ZDR) | 1,048,576 | Text, Images | Tools, Reasoning | Smart | |
| Gemini 3.5 Flash-Lite | EU | None (ZDR) | 1,048,576 | Text, Images | Tools, Reasoning | Swift | |
| Gemini 3.8 Flash | EU | None (ZDR) | 1,048,576 | Text, Images | Tools, Reasoning | Smart | |
| Gemini 3.8 Flash (USA) | USA | None (ZDR) | 1,048,576 | Text, Images | Tools, Reasoning | Smart | |
| Gemini 4 Argon | EU | None (ZDR) | 1,048,576 | Text, Images | Tools, Reasoning | Deep | |
| Gemini 4 Carbon | EU | None (ZDR) | 1,048,576 | Text, Images | Tools, Reasoning | Deep | |
| Gemma 3 27B | EU | None (ZDR) | 128,000 | Text | Tools, Reasoning | Swift | |
| MiniMax M2.5 | MiniMax | EU | None (ZDR) | 128,000 | Text | Tools, Reasoning | Swift |
| Mistral Large 4 | Mistral | EU | None (ZDR) | 524,288 | Text, Images | Tools, Reasoning | Smart |
| Mistral Medium 3.5 | Mistral | EU | None (ZDR) | not stated | not stated | not stated | Smart |
| Mistral Small | Mistral | EU | None (ZDR) | 256,000 | Text, Images | Tools, Reasoning | Swift |
| Kimi K2.6 | Moonshot AI | USA | None (ZDR) | not stated | not stated | not stated | Deep |
| Kimi K3 | Moonshot AI | EU | None (ZDR) | 1,048,576 | Text | Tools, Reasoning | Deep |
| Nemotron 3 Nano Omni | NVIDIA | EU | None (ZDR) | not stated | not stated | not stated | Swift |
| GPT 5.5 | OpenAI | EU | None (ZDR) | 1,050,000 | Text, Images | Tools, Reasoning | Deep |
| GPT 5.6 Luna | OpenAI | USA | None (ZDR) | 1,050,000 | Text, Images | Tools, Reasoning | Smart |
| GPT 5.6 Sol | OpenAI | USA | None (ZDR) | 1,050,000 | Text, Images | Tools, Reasoning | Ultra |
| GPT 5.6 Terra | OpenAI | EU | None (ZDR) | 1,050,000 | Text, Images | Tools, Reasoning | Deep |
| GPT 5.6 Terra (USA) | OpenAI | USA | None (ZDR) | 1,050,000 | Text, Images | Tools, Reasoning | Deep |
| GPT-5 | OpenAI | USA | None (ZDR) | 400,000 | Text, Images | Tools, Reasoning | Smart |
| GPT-5 Mini | OpenAI | EU | None (ZDR) | 200,000 | Text | Tools, Reasoning | Swift |
| GPT-5.4 | OpenAI | EU | None (ZDR) | 1,050,000 | Text, Images | Tools, Reasoning | Smart |
| GPT-6 Astra | OpenAI | USA | None (ZDR) | 1,050,000 | Text, Images | Tools, Reasoning | Ultra |
| GPT-6 Luna | OpenAI | EU | None (ZDR) | 1,050,000 | Text, Images | Tools, Reasoning | Swift |
| GPT-6 Luna (USA) | OpenAI | USA | None (ZDR) | 1,050,000 | Text, Images | Tools, Reasoning | Swift |
| GPT-6 Sol | OpenAI | USA | None (ZDR) | 1,050,000 | Text, Images | Tools, Reasoning | Deep |
| GPT-6.1 Sol | OpenAI | EU | None (ZDR) | 1,050,000 | Text, Images | Tools, Reasoning | Deep |
| GPT-6.1 Sol (USA) | OpenAI | USA | None (ZDR) | 1,050,000 | Text, Images | Tools, Reasoning | Deep |
| GPT-6.1 Terra | OpenAI | EU | None (ZDR) | 1,050,000 | Text, Images | Tools, Reasoning | Deep |
| GPT-6.1 Terra (USA) | OpenAI | USA | None (ZDR) | 1,050,000 | Text, Images | Tools, Reasoning | Deep |
| Perplexity Pro | Perplexity | USA | None (ZDR) | 131,072 | Text | Web search with sources | Swift |
| GLM 5.2 | Z.ai | EU | None (ZDR) | 1,000,000 | Text | Tools, Reasoning | Smart |
| GLM 5.3 | Z.ai | EU | None (ZDR) | 1,000,000 | Text | Tools, Reasoning | Swift |
| Grok 4.5 | xAI | USA | None (ZDR) | 500,000 | Text, Images | Tools, Reasoning | Smart |
| Grok 4.6 | xAI | USA | None (ZDR) | 500,000 | Text, Images | Tools, Reasoning | Smart |
| Grok 4.7 | xAI | USA | None (ZDR) | 500,000 | Text, Images | Tools, Reasoning | Smart |
Context window and maximum output are the operator's figures in tokens. In a chat anymize does not always use the full window and summarizes long histories before it fills up. The class is the usage class from the model picker. Where a cell says “not stated”, we have no verified figure.
Zero data retention
The models in anymize run with zero data retention: the provider does not store prompts or answers, only technical metadata such as time, model and token count. No provider uses the data for training.
The exception is Claude Fable 5, 5.1 and 5.2. For these Anthropic stores prompts and answers for 30 days for abuse detection, and zero data retention is not available for them. With anonymization this only concerns placeholders.
In a team, an admin can block such models with the rule “Only models with zero data retention”. They then show greyed out in the model picker, with the reason next to them.
Retirements
Makers and operators switch older models off at some point. Once an operator has a retirement date, anymize takes it out of routing 7 days before. Your requests then run through the remaining operators in the same jurisdiction, and nobody sees an error on the day.
If no operator is left for a model, it disappears from the picker and its page here is removed.
Alle Seiten in diesem Bereich
Alibaba
Anthropic
DeepSeek
Mistral
Moonshot AI
NVIDIA
OpenAI
Z.ai