Skip to main content
← back to models

// model · coder

Coder model.

The default model on every TartarusAI tier. Custom-tuned MoE architecture — 30B parameters total, 3B active per token, ~17 GB AWQ. Optimized for code generation, refactor, and multi-file edits with low TTFT and high throughput. Same uncensored posture as the rest of the stack.

Architecture
MoE · 30B total · 3B active
Context window
128K (Pro) · 256K (Pro+)
Quantization
AWQ · ~17 GB
Strengths
Code gen, refactor, multi-file edits
Tool calling
Yes · OpenAI tool_calls + Llama-style fallback
Reasoning trace
No · use Thinking model

Coder is the right default for ~80% of workflows. Switch to Thinking when you need visible reasoning, multi-step planning, or architectural review.