Model Tiers

Cognova uses Anthropic's Claude models exclusively. We offer three tiers optimized for different use cases.

Tiers

Fast — Claude Haiku 4.5

• Best for: Quick lookups, simple Q&A, classification, summarization
• Speed: Fastest response times
• Available on: All plans

Haiku is ideal for high-volume, low-complexity tasks. Use it for agents that handle routine queries, format data, or triage incoming requests.

Balanced — Claude Sonnet 4.6

• Best for: Code generation, analysis, detailed explanations, multi-step reasoning
• Speed: Fast with significantly higher capability
• Available on: Starter and above

Sonnet is the best default for most agents. It handles complex reasoning, code review, and nuanced analysis while keeping response times reasonable.

Pro — Claude Opus 4.6

• Best for: Complex architecture decisions, deep analysis, creative writing, research
• Speed: Slower but highest capability
• Available on: Pro and above

Opus is the most capable model. Use it for agents that need maximum intelligence — system design, comprehensive code review, or tasks where quality matters more than speed.

Choosing a Tier

Setting Model Tier

Each agent has its own model tier setting:

1. Go to Agents in the sidebar
2. Click the agent to edit
3. Change the Model Tier dropdown
4. Save

The model tier only affects that specific agent. You can mix tiers across agents — use Haiku for your triage agent and Opus for your code reviewer.

Credit Costs

Each message costs credits based on the model tier:

Tool calls, web searches, and agent delegation within a single message are included at no extra cost. Detailed usage (including per-agent breakdown and API costs) is visible in Settings > Billing.