Coordinate coding agents.

Torad Fleet keeps native coding-agent sessions visible, assigns ownership, and prevents overlapping edits.

Compact-model research

Current Torad runs measure memory, throughput, and training on consumer hardware. Independent reproduction is still pending.

Qwen3-1.7B linear weights

870 MB

TQ4 used 870 MB. The bf16 comparison used 2.7 GB.

TQ4 runtime

238 tok/s

vLLM reached 267 tok/s in the comparison.

Gemma 4 E2B training

13.5 GB

Peak use on an RTX 5080 with 16 GB.