Compare models

kunji compare sends one prompt to several models at the same time, shows each answer, and ends with a summary table.

kunji compare "Write a regex that matches email addresses" \
  -m openai:gpt-4o-mini -m groq:llama-3.3-70b-versatile -m anthropic:claude-sonnet-4-5

✓ openai:gpt-4o-mini  2.1s · 96 tok
✓ groq:llama-3.3-70b-versatile  0.6s · 88 tok
✓ anthropic:claude-sonnet-4-5  3.0s · 120 tok

┌─  Compare  ──────────────────────────────────────────────────────────┐
│  Model                            Time   Tokens      Cost            │
│──────────────────────────────────────────────────────────────────────│
│  groq:llama-3.3-70b-versatile     0.6s       88   <$0.0001   fastest │
│  openai:gpt-4o-mini               2.1s       96    $0.0001  cheapest │
│  anthropic:claude-sonnet-4-5      3.0s      120    $0.0019           │
└──────────────────────────────────────────────────────────────────────┘

(Example output; your times and prices will differ.)

Without choosing models

kunji compare "Explain recursion"

uses your default model plus the default models of your other providers (up to three).

From the menu

Compare models → type the prompt → tick models with Space → Enter.

Good uses

• Pick the cheapest model that’s good enough for a task you do often.
• Check an important answer against a second model.
• See how fast a new provider really is.

Every answer is billed by its provider and counted in kunji usage as normal.