Valstorm Logo
ValstormDocs
AI and Agents•v1.0.0•3 min read

AI Agent Model Strategy & Allocation Specification

Three-tier cognitive model allocation strategy across heavy thinking architects, reasoning orchestrators, and high-throughput worker sub-agents.

The Valstorm Agent Runtime utilizes a Three-Tier Cognitive Model Architecture that optimizes reasoning depth, latency, and token economics across multi-agent workflows.

1. Executive Summary & Model Tier Taxonomy

┌────────────────────────────────────────────────────────────────────────┐ │ TIER 1: HEAVY THINKERS │ │ System Architecture • Technical Planning • High-Stakes Ideation │ │ Gemini 3.1 Pro Preview | Claude Opus 5 | gpt-5.6-sol │ └───────────────────────────────────┬────────────────────────────────────┘ │ Generates Architecture & Specs ▼ ┌────────────────────────────────────────────────────────────────────────┐ │ TIER 2: REASONERS │ │ Orchestration • Monorepo Dev • Tool Calling • Pytest & Playwright │ │ Gemini 3.7 / 3.6 Flash | Claude Sonnet 5 | gpt-5.3-codex │ └───────────────────────────────────┬────────────────────────────────────┘ │ Delegates Micro-Tasks ▼ ┌────────────────────────────────────────────────────────────────────────┐ │ TIER 3: WORKER BEES │ │ Spec Execution • CUD Mutations • Schema Lookups • Data Scraping │ │ Gemini 3.5 / 3.1 Flash-Lite | Claude Haiku 4.5 | gpt-5.6-luna │ └────────────────────────────────────────────────────────────────────────┘

2. Model Tier Specification & Best Use Cases

Tier 1: Heavy Thinkers (Architects & High-Stakes Planners)

  • Role: High-level system architecture, multi-file code planning, complex database schema design, technical trade-off evaluation, and deep root-cause debugging.
  • When to Use: Invoked during the initial architecture phase or when an escalation threshold is reached.
  • Recommended Models:
    • Google Gemini: gemini-3.1-pro-preview — Massive context window, superior structural reasoning.
    • Anthropic Claude: claude-opus-5 — Nuanced comprehension and architectural prose.
    • OpenAI: gpt-5.6-sol / gpt-5.5 — Deep chain-of-thought and mathematical rigor.

Tier 2: Reasoners (Orchestrators, Full-Stack Devs & QA Engineers)

  • Role: Interactive REPL orchestration, multi-tool agent execution, monorepo coding (FastAPI + React), test suite authoring (Pytest/Playwright), and dynamic debugging.
  • When to Use: Primary engine for interactive turns and specialized agent profiles.
  • Recommended Models:
    • Google Gemini: gemini-3.7-flash / gemini-3.6-flash — High speed-to-intelligence ratio, reliable function calling.
    • Anthropic Claude: claude-sonnet-5 — Precise code generation and refactoring.
    • OpenAI: gpt-5.6-terra / gpt-5.3-codex — Specialized code completions and fast reasoning.

Tier 3: Worker Bees (Spec Executors & Single-Task Micro-Agents)

  • Role: High-throughput execution of rigid specifications, single-field CRUD mutations, schema lookups, basic SQL generation, log aggregation, and data formatting.
  • When to Use: High-throughput automated pipelines and sub-tasks delegated by parent agents.
  • Recommended Models:
    • Google Gemini: gemini-3.5-flash-lite / gemini-3.1-flash-lite — Ultra-low cost, sub-second latency.
    • Anthropic Claude: claude-haiku-4.5 — Fast, structured JSON generation.
    • OpenAI: gpt-5.6-luna / gpt-5.4-nano — Lightweight subagent execution.

3. Subagent Task Delegation Pipeline

When an orchestrating agent invokes delegate_task(profile="researcher", goal="..."), the runtime executes the child agent in an isolated context window with its allocated model tier. The child synthesizes the output and returns a concise report to the parent, preserving the primary agent's context budget.