Models
The models the router can choose from.
Local models are discovered from Ollama; cloud models come from the configured catalog. Every model sits behind the same provider interface, and capability profiles are configuration, not benchmarks.
Local models are discovered from Ollama; cloud models come from the configured catalog. Every model sits behind the same provider interface, and capability profiles are configuration, not benchmarks.
Routing configuration, not independent benchmarks. Tune them with evidence from the Comparison Lab.
Owner's local eval 2026-10-04 (thinking off): 5/5 graded prompts, 5/5 on 10K-token retrieval. Speeds from the same benchmark.
Owner's local eval 2026-10-04 (thinking off): missed one arithmetic step, a word limit and one long-context lookup; ~30% faster generation than Gemma 4 12B.
Economical-tier default. Routing configuration, not a benchmark.
General high-quality tier default. Routing configuration, not a benchmark.
Premium reasoning tier default. Routing configuration, not a benchmark.
As installed on the developer's Mac. Execution is simulated on this site.
| Model | Parameters | Size | Context (effective / max) | Throughput | Status | Routable |
|---|---|---|---|---|---|---|
Gemma 4 12B gemma4:12b-it-qat | 11.9B Q4_0 | 7.2 GB | 32K / 256K | 15.9 tok/s configured | Simulated | Yes |
Qwen 3.5 9B qwen3.5:9b | 9.0B Q4_K_M | 6.6 GB | 32K / 256K | 21.1 tok/s configured | Simulated | Yes |
ollama list and runs them for real. Throughput figures are the developer's own measurements on an M3 Pro.List prices per million tokens. Capability values are routing configuration.
| Model | Tier | Input / MTok | Output / MTok | Context | Highest data level | Execution | Routable |
|---|---|---|---|---|---|---|---|
Claude Haiku 5.5 Anthropic · claude-haiku-5-5 | economical | $0.10 | $0.50 | 1M | confidential | Simulated | Yes |
Claude Sonnet 5.5 Anthropic · claude-sonnet-5-5 | general | $2.00 | $10.00 | 1M | confidential | Simulated | Yes |
Claude Opus 5.5 Anthropic · claude-opus-5-5 | premium | $4.00 | $20.00 | 1M | confidential | Simulated | Yes |
OpenAI Adapter implemented, not configured | Available on a local install once configured. No prices are guessed. | ||||||
Google Gemini Adapter implemented, not configured | Available on a local install once configured. No prices are guessed. | ||||||