Supported presets

Aspose.LLM for .NET uses presets to configure the model, context, chat, and sampler parameters. You pass a preset to AsposeLLMApi.Create(preset). The base type is PresetCoreBase (from the Aspose.LLM.Abstractions package).

Text presets

Built-in presets for text-only models include:

• Qwen25Preset — Qwen 2.5 7B Instruct (GGUF)
• Qwen3Preset — Qwen 3
• Gemma3Preset — Gemma 3
• Llama32Preset — Llama 3.2
• Phi4Preset — Phi 4
• DeepseekR1Qwen3Preset — DeepSeek R1 / Qwen 3
• DeepSeekCoder2Preset — DeepSeek Coder 2
• Oss20Preset — OSS 2.0

Vision presets

For models that support images (multimodal input), use a vision preset:

• Qwen25VL3BPreset — Qwen 2.5 VL (vision-language) 3B
• Qwen3VL2BPreset — Qwen 3 VL 2B
• Gemma3VisionPreset — Gemma 3 Vision
• Ministral3VisionPreset — Ministral 3 Vision

Use the optional media parameter of SendMessageAsync or SendMessageToSessionAsync to pass image (or other) bytes when using a vision preset.

Default preset

GetDefaultPreset() returns a preset with engine default parameters (e.g. Qwen25Preset). You can also obtain raw default parameters with GetDefaultParametersAsync().