AI Models

DevOS supports 40+ AI models across 5 providers. Use the best model for each task, or let Smart Router choose automatically.

OpenAI

Model	Best For	Cost (1M tokens)
`gpt-5.4-pro`	Most capable (recommended)	$5 / $20
`gpt-5.4`	General coding	$4 / $16
`gpt-5-turbo`	Fast & capable	$3 / $12
`gpt-4o`	Balanced cost/quality	$2.50 / $10
`gpt-4o-mini`	Fast tasks, low cost	$0.15 / $0.60
`o3`	Advanced reasoning	$10 / $40
`o4-mini`	Fast reasoning	$1.10 / $4.40

devos config set OPENAI_API_KEY=sk-your-key
devos model use gpt-5.4-pro

Anthropic (Claude)

Model	Best For	Cost (1M tokens)
`claude-4.7-sonnet-20260401`	Best coding model (recommended)	$4 / $20
`claude-4.7-opus-20260401`	Most capable, complex analysis	$20 / $100
`claude-4.6-opus-20260201`	Premium reasoning	$18 / $90
`claude-4.6-sonnet-20260201`	Balanced performance	$3.50 / $18
`claude-sonnet-4-20250514`	Previous gen, still great	$3 / $15
`claude-3-5-haiku-20241022`	Fast, affordable	$0.80 / $4

devos config set ANTHROPIC_API_KEY=sk-ant-your-key
devos model use claude-4.7-sonnet-20260401

Google Gemini

Model	Best For	Cost (1M tokens)
`gemini-3.1-pro`	Most capable (recommended)	$2 / $12
`gemini-3.1-flash`	Ultra fast & cheap	$0.20 / $0.80
`gemini-2.5-pro`	Previous gen, solid	$1.25 / $10
`gemini-2.5-flash`	Budget option	$0.15 / $0.60

devos config set GOOGLE_API_KEY=AIza-your-key
devos model use gemini-3.1-pro

Ollama (Free, Local)

Run models completely offline on your machine. No API keys, no costs.

Model	Best For	RAM Required
`llama3.3`	General coding	~8 GB
`codellama`	Code generation	~8 GB
`deepseek-coder-v2`	Code + reasoning	~16 GB
`mistral`	Fast general use	~4 GB
`qwen2.5-coder`	Code specialist	~8 GB

# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# Pull a model
ollama pull llama3.3

# Use in DevOS
devos model add ollama
devos model use llama3.3

OpenRouter (25+ Models)

Access models from every provider through a single API key.

devos config set OPENROUTER_API_KEY=sk-or-your-key
devos model use openrouter/anthropic/claude-sonnet-4-20250514

Smart Router

Smart Router automatically selects the best model based on task complexity:

Low complexity (simple edits, explanations) → Fast/cheap model (GPT-4o-mini, Gemini 3.1 Flash)
Medium complexity (feature development) → Balanced model (GPT-5.4 Pro, Claude 4.7 Sonnet)
High complexity (architecture, debugging) → Premium model (Claude 4.7 Opus, Gemini 3.1 Pro)

# Enable Smart Router
devos model auto

💡 Cost Savings Smart Router can reduce AI costs by 60-80% by using cheaper models for simple tasks while reserving expensive models for complex operations.

Cost Tracking

Track your AI spending in real-time:

devos cost          # View current costs
devos cost --reset  # Reset counters

DevOS tracks tokens used and calculates costs per-provider with accurate pricing tables for all supported models.