The Gemini Planet Imager (GPI), a coronagraphic adaptive optics instrument designed for spectroscopy of extrasolar planets, had first light in 2013 [13]. After five years, GPI has observed more than 500 stars, producing an extensive library of science images and associated telemetry that can be analyzed to determine performance predictors. We will present a summary of on-sky performance and lessons learned. The two most significant factors determining bright star contrast performance are atmospheric coherence time and the presence of dome seeing. With a possible move to Gemini North, we are planning potential upgrades including a pyramid-sensor based AO system with predictive control; we will summarize upgrade options and the science they would enable.

The Gemini Planet Imager (GPI) is the dedicated high-contrast imaging facility, located on Gemini South, designed for the direct detection and characterization of young Jupiter mass exoplanets. In 2019, Gemini is considering moving GPI from Gemini South to Gemini North. Analysis of GPI's as-built performance has highlighted several key areas of improvement to its detection capabilities while leveraging its current capabilities as a facility class instrument. We present the proposed upgrades which include a pyramid wavefront sensor, broadband low spectral resolution prisms and new apodized-pupil Lyot coronagraph designs all of which will enhance the current science capabilities while enabling new science programs.An active area of study for young, giant exoplanets is how the current luminosity of these young planets traces their formation mechanisms, and whether these planets can be described by a "hot start" luminosity model (likely corresponding to a disk instability formation mechanism), or the alternate "cold start" models, which is expected from cold start formation. The current GPIES survey from Gemini South is quite sensitive to hot-start giant planets between 10-100 AU, but the survey sensitivity is comparatively low to cold-start planets (Figure 1). Indeed, based on current cold-start evolutionary models 34 coupled with AMES-COND atmosphere models, 35,36