A Pi extension that replaces the basic model picker with a searchable /models picker built for choosing coding models.
- Artificial Analysis Coding Index score and letter grade (
A+toF) - Local rank among the models available in your current Pi session
- Routing provider shown for every model
- Input/output pricing
- Optional context-window display
- Fuzzy search
Tab/Shift+Tabmodel cyclingShift+←/Shift+→thinking-level control- Full-width selected-row highlighting
/modelscommand plusCtrl+Lmodel-picker integration
The package is not published on npm yet. Use the GitHub install for now:
pi install git:github.com/ktappdev/pi-better-modelsOnce published, the npm install will be:
pi install npm:@kentaylorappdev/pi-better-modelsNPM package page · GitHub repository
The picker uses the Artificial Analysis Coding Index, a 0–100 coding-focused score:
- It first reads
benchmarks.artificial_analysis.codingfrom modelgrep. - If modelgrep has no coding score, it uses the first-party Artificial Analysis API when an API key is configured.
- If neither source has a coding score, the model is shown as unscored.
There is no intelligence-score rescaling or custom heuristic. The displayed rank is calculated locally across your available, scored models.
Grades are calibrated to the current Coding Index range:
80+ A+ 76–79 A 73–75 A− 70–72 B+
67–69 B 64–66 B− 61–63 C+ 58–60 C
55–57 C− 50–54 D <50 F
Modelgrep works without a key. To fill coding-score gaps with Artificial Analysis’s first-party data, set a free API key:
export ARTIFICIAL_ANALYSIS_API_KEY=aa_xxxAA_API_KEY is also accepted as an alias. Restart Pi after changing the environment.
The default compact row shows pricing and rating:
pricing · coding rank & score (AA)
Use PI_MODELS_COLUMNS to choose from context, pricing, and score:
# Default
export PI_MODELS_COLUMNS=pricing,score
# Show all fields
export PI_MODELS_COLUMNS=context,pricing,score
# Context and rating only
export PI_MODELS_COLUMNS=context,scoreMIT © Ken Taylor