The Pass Preset

The Current column is the house preset in use — retuned on cache-free measurements, balancing speed and price. Provider columns list interchangeable equivalents. Version 2026-08-19d, prices input/output USD per 1M tokens.

Task class	Current (house)	Google AI	OpenAI (native)	Anthropic	Fireworks	OpenRouter
Small — the Desk's brain	OpenRouter `gemini-2.5-flash-lite` $0.1/$0.4 · 2.2s	`gemini-2.5-flash-lite` $0.1/$0.4 backup: gemini-3.1-flash-lite	`gpt-5.4-nano` $0.2/$1.25 backup: gpt-4o-mini	`claude-haiku-4-5` backup: claude-3-haiku	`gpt-oss-20b`	`gemini-2.5-flash-lite` $0.1/$0.4 backup: gpt-oss-20b
Default — everyday asks	Fireworks `minimax-m2p7` · 3.9s	`gemini-3.5-flash` $1.5/$9 backup: gemini-2.5-flash	`gpt-5.4-mini` $0.75/$4.5 backup: gpt-5-mini	`claude-haiku-4-5` backup: claude-sonnet-5	`minimax-m2p7` backup: gpt-oss-20b	`gpt-oss-20b` $0.03/$0.13 backup: qwen3-30b-a3b-instruct-2507
Code	OpenRouter `codestral-2508` $0.3/$0.9 · 5.1s	`gemini-3.5-flash` $1.5/$9 backup: gemini-2.5-pro	`gpt-5.2-codex` $1.75/$14 backup: gpt-5.1-codex	`claude-sonnet-5` $2/$10 backup: claude-haiku-4-5	`minimax-m3`	`codestral-2508` $0.3/$0.9 backup: qwen3-coder-30b-a3b-instruct
Long-form writing	Fireworks `minimax-m3` · 6.8s	`gemini-2.5-pro` $1.25/$10 backup: gemini-3.5-flash	`gpt-5.2` $1.75/$14 backup: gpt-5.1	`claude-sonnet-5` $2/$10 backup: claude-haiku-4-5	`minimax-m3`	`mistral-large-2512` $0.5/$1.5 backup: qwen3.7-plus
Analysis & reasoning	OpenAI `gpt-4o` $2.5/$10 · 8.0s	`gemini-2.5-pro` $1.25/$10 backup: gemini-3.5-flash	`gpt-5.2` $1.75/$14 backup: o4-mini	`claude-sonnet-5` $2/$10 backup: claude-opus-5	`nemotron-3-ultra-nvfp4`	`deepseek-v3.2` $0.269/$0.4 backup: grok-4.20

Names shown without the provider path; full ids in preset.json. Analysis uses a faster-but-pricier model deliberately (the cheaper options ran 18–39 s). Overridable in Desk settings.