Skip to content
AMC WebUI

Model Matrix

Built-in Model Matrix

AMC WebUI natively supports Google’s latest Gemini 3.x series, embodied robotics models, and dedicated multimodal endpoints.

Built-in Native Gemini Model Matrix

Gemini 3.8 FlashRecommended
gemini-3.8-flash

Default powerhouse model. Fast responses, strong reasoning, and multimodal capabilities.

1M Tokens深度思考 (Low~High)搜索与地图增强代码执行Live Artifacts
Gemini 3.5 Flash-Lite
gemini-3.5-flash-lite

Lightweight efficient model used for real-time thinking process translation and quick tasks.

1M Tokens极速思考 (Minimal起)思维链实时翻译极低延迟
Gemini 3.1 Pro Preview
gemini-3.1-pro-preview

High-capability model with 2M context and up to 32K thinking budget for complex math and coding.

2M Tokens高阶数理超大上下文32K Token 思考预算
Gemini Robotics-ER 2
gemini-robotics-er-2-preview

Embodied spatial reasoning model requiring API key domain/IP restrictions in Google AI Studio.

1M Tokens具身空间推理需绑定 API 限制建议思考 Medium
Gemini 3.1 Flash Live
gemini-3.1-flash-live-preview

Dedicated Live API endpoint for bidirectional audio/video chat with sub-second latency.

128K Tokens双向流式语音摄像头识别屏幕共享理解AudioWorklet
Gemini 3 Pro Image (Nano Banana)
gemini-3-pro-image-preview

Native high-quality text-to-image model supporting custom aspect ratios and multi-image generation.

32K Tokens原生生图比例定制 (1:1 / 16:9)四图并发
Gemini 3.1 Flash TTS
gemini-3.1-flash-tts-preview

Expressive speech synthesis with 30 lifelike voices and real-time speed adjustment.

32K Tokens30 种拟真音色语速动态微调多语种自然重音

Model Selection Guidelines

1. Primary Workhorse: gemini-3.8-flash

  • Role: Default recommended model for daily programming, writing, tool invocation, and long-context processing.
  • Highlights: Extremely low latency, full Thinking levels (Low / Medium / High), multimodal file analysis, and Live Artifacts generation.

2. Ultra-fast & Lightweight: gemini-3.5-flash-lite

  • Role: High-throughput micro-queries, rapid classifications, and the dedicated background engine for real-time chain-of-thought translation.
  • Highlights: Minimal token cost and latency with support for the Minimal thinking level.

3. Advanced Reasoning: gemini-3.1-pro-preview

  • Role: Complex mathematical reasoning, academic research, and massive cross-repository codebase refactoring.
  • Highlights: 2M token context window and up to 32,768 thinking token budget.

4. Embodied & Spatial Intelligence: gemini-robotics-er-2-preview

  • Role: Spatial geometry, 3D spatial relationship deduction, and robotic task planning.
  • Key Requirement:
    • Google requires the calling API Key to have API Restrictions applied in Google AI Studio. Unrestricted keys will be rejected with 403 Forbidden.
    • Recommended Thinking Level: Medium to balance spatial calculation accuracy with latency.