Legacy build│
Firmware 1.2
Switched to quantized model formats to halve boot time and reduce RAM footprint on 8 GB machines.
// Highlights
Quantized Models
Models swapped to quantized variants — comparable quality, half the RAM, faster first-token latency.
// What changed
- Quantized model pipeline (Q4/Q5 variants).
- Cold-boot time reduced by ~50% on typical hardware.
- 8 GB RAM machines now run comfortably with the standard preset.
// Hardening
- Memory pressure handling for sustained chat sessions.
- Cleaner shutdown when USB is ejected unexpectedly.
← Previous
Firmware 1.1
Added the system prompt editor and basic thread persistence to the desktop UI.
Next →
Firmware 1.3
Added official macOS launcher (Intel + Apple Silicon) and the offline preset library with multiple speed/depth profiles.