Model Catalog

Meet lucif.

Not a single frozen model. A living meta-model: lucif continuously routes to the most optimal model for agentic & coding workloads, and the fleet behind it is refreshed as new models are released, inspired by the "one best endpoint" philosophy of modern routing research.

LUCIF ROUTING SIMULATION your-agent lucif auto-route · failover ✗ claude-opus-5 ✓ HEALTH 100 · P50 212MS kimi-k3 · STANDBY · BENCH 98 deepseek-v4-pro ✗ DEGRADED · SKIPPED live routing · unhealthy lanes are skipped automatically · no config needed
Flagship
lucif

The default endpoint. Every call is scored against live benchmarks and routed to the model with the best quality × latency × cost balance for agentic and coding tasks.

60 RPM · Unlimited Token
CLI-compatible alias
lucif-claude

Identical routing core, identical performance. The -claude suffix exists purely so CLIs that require the "claude" initials in a model name, like Claude Code Desktop App, can connect without friction.

60 RPM · Unlimited Token
The value proposition

Why "most optimal" beats "most expensive"

Poor quality & latency Huge costs, marginal gains Low-Budget Model Flagship Overkill LUCIF · Most Optimal Model Best value between quality and gains
Live models

What we provide

The models currently powering Lucif-automatically route requests to the best performing and most up-to-date models available in real time. We strive to provide only models with high-quality reasoning capabilities for agentic tasks and coding.