Model Selection Guide

Available Models Overview

The Knowledge Assistant supports multiple large language models (LLMs) from different providers. Each model has different characteristics in terms of speed, quality, and cost. You can select your preferred model from the settings menu.

Google Models (Gemini)

Gemini 3 Flash

Recommended

Google's latest fast model optimized for speed while maintaining high quality. Excellent balance of performance and response time for most use cases.

Speed: Very fast (~1-2s TTFT)
Quality: High
Knowledge+: Supported

Gemini 3 Pro

Google's most capable reasoning model with enhanced analytical abilities. Better for complex questions requiring multi-step reasoning.

Speed: Moderate (~2-4s TTFT)
Quality: Excellent
Knowledge+: Supported

Anthropic Models (Claude)

Claude Sonnet 4.5

Anthropic's balanced model with strong reasoning and writing capabilities. Excellent for nuanced analysis and detailed explanations.

Speed: Moderate (~2-3s TTFT)
Quality: Excellent
Knowledge+: Not supported (requires Google Search)

Claude Haiku 4.5

Anthropic's fastest model, optimized for quick responses while maintaining good quality. Ideal for simple queries and high-throughput applications.

Speed: Very fast (~1-2s TTFT)
Quality: Good
Knowledge+: Not supported (requires Google Search)

Speed vs. Quality Trade-offs

ModelSpeedQualityBest For
Gemini 3 Flash★★★★★★★★★☆Daily use, most queries
Gemini 3 Pro★★★☆☆★★★★★Complex analysis, reasoning
Claude Sonnet 4.5★★★☆☆★★★★★Detailed writing, nuance
Claude Haiku 4.5★★★★★★★★☆☆Quick lookups, simple queries

Cost Implications

Different models have different costs per token. While the Knowledge Assistant manages this internally, understanding the relative costs can help you choose:

Relative Cost Comparison

Gemini 3 Flash / Claude Haiku 4.5$
Claude Sonnet 4.5$$
Gemini 3 Pro$$$

Knowledge+ Mode Restrictions

Google Models Required for Knowledge+

Knowledge+ mode uses Google Search grounding, which is only available through Google's Gemini models. If you switch to Knowledge+ mode while using a Claude model, the system will automatically switch to Gemini Pro (default). Gemini Flash is used only when you explicitly request a fast model.

Supported in Knowledge+: Gemini 3 Flash, Gemini 3 Pro

When to Use Each Model

Use Gemini 3 Flash when:

  • You need fast responses
  • Queries are straightforward
  • You're using Knowledge+ mode and want the fast path
  • Cost efficiency matters

Use Gemini 3 Pro when:

  • Questions require complex reasoning
  • You need analytical depth
  • You want the default Knowledge+ model
  • You're using Knowledge+ mode with complex queries
  • Quality is more important than speed

Use Claude Sonnet 4.5 when:

  • You need nuanced, detailed writing
  • Questions have multiple perspectives
  • You're in Bitwise Knowledge mode only
  • Response quality is paramount

Use Claude Haiku 4.5 when:

  • Speed is critical
  • Queries are simple lookups
  • High volume of questions
  • Cost efficiency is important

Related Articles