Models and providers

All Models and Pricing

5 providers, 15 models, from free to frontier. BYOK — you pay providers directly through your own API keys.

Model Reference (per 1M tokens)

---

Judge Model Priority

The judge evaluates proposals and synthesizes the final answer. Selected by priority:

#1 Anthropic → #2 Google → #3 OpenAI → #4 xAI → #5 Groq

Uses the first provider for which you have a valid key.

---

Free Fallback System

When no paid API keys are configured, Consilium automatically falls back to Groq's free tier:

• llama-3.1-8b-instant (fast, 8B params)
• llama-3.3-70b-versatile (balanced, 70B params)
• llama-4-scout-17b (latest, 17B params)

Zero cost for prototyping and testing.

---

Context Overflow Fallback

When a model returns 413/400 (context too large), Consilium retries with a cheaper variant:

---

Provider Environment Variables

All keys encrypted with AES-256-GCM before storage. Never stored in plaintext. Never logged.