// SYSTEM_ARCHITECTURE v3
Greg's AI Workflow
How requests flow through the agent stack
INPUT
User Request
O
Opus 4.6
PLANNER
Reasoning, strategy, architecture, plans
native
S
Sonnet 4.6
LEAD AGENT
Primary executor · Reviews all outputs · Integrates · Deploys
native
delegates via MCP
RESEARCH / REASONING
Gemini 2.5 Flash
Research, blog posts, FAQs, comparisons, case studies
research
compare
find_best
Gemini 2.5 Pro
PREMIUM
Deep reasoning, complex analysis, architecture decisions
pro_reason
deep_research
Gemini 3.1 Pro
FRONTIER
Most capable model — hardest problems, deepest analysis
frontier
Groq
5 models
Llama 3.3 70B · Llama 4 Maverick · Scout · Llama 3.1 8B · Whisper v3
groq-fast
fast_transcribe
CODE / BACKEND
GPT 5.1 Codex
Complex backend, algorithms, APIs
opencode_gpt51_codex
Kimi K2 Thinking
Complex logic, multi-step reasoning
opencode_kimi_thinking
Trinity Large
General coding and reasoning
opencode_trinity
GLM 5
FREE
General coding, analysis
opencode_glm5_free
CONTENT / UTILITY
Kimi K2.5
FREE
Bulk HTML generation, templates
opencode_kimi_free
GLM 4.7
FREE
Chat, writing, analysis, code review
glm_chat
glm_code
glm_write
Ollama
Llama 3.2 3B
LOCAL
Summaries, rewrites, code review — unlimited
ollama-local
Results flow back to Sonnet 4.6 for review, integration & deployment
// REFERENCE_TABLE
| Role | Model | MCP Server | Use Case | Cost |
|---|---|---|---|---|
| Planner | Opus 4.6 | native | Reasoning, architecture, strategy | paid |
| Lead | Sonnet 4.6 | native | Executes tasks, reviews, deploys | paid |
| Research | Gemini 2.5 Flash | gemini-research | Research, comparisons, blog posts, FAQs | paid |
| Reasoning | Gemini 2.5 Pro | gemini-research | Deep analysis, architecture, complex reasoning | paid |
| Frontier | Gemini 3.1 Pro | gemini-research | Hardest problems, frontier intelligence | paid |
| Fast | Llama 3.3 70B | groq-fast | Quick answers, analysis, code snippets | FREE |
| Fast | Llama 4 Maverick 17B | groq-fast | 128-expert MoE, complex tasks | FREE |
| Fast | Llama 4 Scout 17B | groq-fast | 16-expert MoE, lighter & faster | FREE |
| Fast | Llama 3.1 8B | groq-fast | Ultra-fast small model, instant answers | FREE |
| Audio | Whisper v3 Turbo | groq-fast | Speech-to-text transcription | FREE |
| Backend | GPT 5.1 Codex | opencode | Complex algorithms, APIs | paid |
| Reason | Kimi K2 Thinking | opencode | Complex logic, multi-step planning | paid |
| Code | Trinity Large | opencode | General coding and reasoning | paid |
| Code | GLM 5 | opencode | General coding, analysis | FREE |
| HTML | Kimi K2.5 | opencode | Bulk HTML, templates | FREE |
| General | GLM 4.7 | glm-free | Chat, write, analyze, code review | FREE |
| Local | Llama 3.2 3B | ollama-local | Summaries, rewrites — unlimited, on-device | LOCAL |
17
Models
3
Gemini Tiers
9
Free / Local
5
MCP Servers
28
Total Tools