Coming soon.

Limits

• Global budget: $5 per day for the whole comparator; the maximum cost of each call is reserved before the call, beyond that the comparator refuses.
• 5 model answers per day per visitor (each model queried counts as one; UTC day).
• Message of 500 characters at most; no message is kept.

Models, endpoint and region

• Claude (eu.anthropic.claude-haiku-4-5-20251001-v1:0) — AWS Bedrock, EU inference profile (LLM_PROVIDER=BEDROCK_EU) — eu-west-3 (Paris) (EU)
• GPT (gpt-4o) — api.openai.com — outside the EU — OpenAI (EU data residency not enabled) (outside the EU)
• Gemini (gemini-2.5-flash) — Vertex AI (europe-west1-aiplatform.googleapis.com) — europe-west1 (Belgium) (EU)
• Mistral (mistral-large-latest) — api.mistral.ai — EU as declared by Mistral AI (Paris), not verifiable from the code (EU)