Every AI model you can choose, tested in production.
31 AI models from seven providers, plus models you run on your own hardware. Give each department a different one. On managed plans there is no key to add: pick a model and use it. Every hosted model below answered a test through NearSync AI in October 2026.
Part of NearSync AI: ERP software in Dubai and the UAE.
Supported AI models
Choose any model, per department.
These are the AI models you can choose in NearSync AI today, grouped by the company that makes them. Set one as the base model, then give any department its own: deep reasoning for Finance, something quick for Support. How choosing a model works.
Google Gemini
7 models| Model | Good for | Speed | Context |
|---|---|---|---|
| Gemini 3.8 FlashSuggested | Images and documents, Reasoning | Fast | 1M |
| Gemini 3.5 Flash Lite | Quick answers | Fastest | 1M |
| Gemini 3.5 Flash | Images and documents, Reasoning | Fast | 1M |
| Gemini 3.1 Pro | Reasoning, Code | Thorough | 1M |
| Gemini 3.1 Flash LiteOlder generation | Quick answers | Fastest | 1M |
| Gemini 2.5 FlashOlder generation | Images and documents | Fast | 1M |
| Gemini 2.5 ProOlder generation | Reasoning | Thorough | 1M |
OpenAI
8 models| Model | Good for | Speed | Context |
|---|---|---|---|
| GPT-5.6 TerraSuggested | Images and documents, Reasoning | Balanced | 1M |
| GPT-5.6 Luna | Quick answers | Fast | 1M |
| GPT-5.6 Sol | Reasoning, Code | Thorough | 1M |
| GPT-6 Astra | Reasoning, Code | Thorough | 1M |
| GPT-5.4Older generation | Images and documents, Reasoning | Balanced | 128K |
| GPT-5.4 MiniOlder generation | Quick answers | Fast | 128K |
| GPT-5.4 NanoOlder generation | Quick answers | Fastest | 128K |
| o3-miniOlder generation | Reasoning, Code | Thorough | 200K |
Anthropic Claude
6 models| Model | Good for | Speed | Context |
|---|---|---|---|
| Claude Sonnet 5Suggested | Code, Analysis | Balanced | 1M |
| Claude Haiku 4.5 | Quick answers | Fastest | 200K |
| Claude Opus 5 | Reasoning, Code | Thorough | 1M |
| Claude Fable 5.1 | Reasoning, Code | Thorough | 1M |
| Claude Sonnet 4.6Older generation | Code, Analysis | Balanced | 200K |
| Claude Opus 4.8Older generation | Reasoning | Thorough | 200K |
xAI Grok
3 models| Model | Good for | Speed | Context |
|---|---|---|---|
| Grok-3 Mini FastSuggested | Quick answers | Fastest | 128K |
| Grok-3 Mini | Reasoning | Balanced | 128K |
| Grok-3 | Reasoning | Thorough | 128K |
Mistral AI
3 models| Model | Good for | Speed | Context |
|---|---|---|---|
| Mistral LargeSuggested | General work | Balanced | 128K |
| Mistral Small | General work | Fast | 32K |
| Codestral | Code | Fast | 32K |
Groq
2 models| Model | Good for | Speed | Context |
|---|---|---|---|
| GPT-OSS 120BSuggested | Open weights, Reasoning | Fastest | 128K |
| GPT-OSS 20B | Quick answers, Open weights | Fastest | 128K |
DeepSeek
2 models| Model | Good for | Speed | Context |
|---|---|---|---|
| DeepSeek V3Suggested | Code | Balanced | 64K |
| DeepSeek R1 | Reasoning, Code | Thorough | 64K |
On your own hardware
Run a model with Ollama on machines you own, for example Llama 3.2, Mistral 7B or Code Llama, and point a department at it. Nothing about that department's work leaves your network.
Kept current
When a provider retires a model, it leaves this list.
AI companies retire models often. This list comes straight from the models NearSync AI offers in the product, and every hosted model on it was tested in October 2026. If a model you use is retired, your department moves to the provider's recommended replacement, and you can change it from one screen.