I carried out a measurement over the weekend comparing the BME680 (eCO2 eq.) to an accurate reference device which directly measures CO2 (Air Control 5000). As one can see in the result diagram the BME680 shows in comparison to the reference device about +100-150 ppm deviation and unrealistic jumps.
I am using the BME680 on an embedded device over I2C with an A20 Allwinner processor (ARM) running Linux (Debian). The code base is taken from this repository: