Same soul, carried forward. Small, fast, fully local.
Genesys e3 is a light fine-tune of Genesys e · 4B — a 4B Genesys model focused on identity, coherence, personality and conversational continuity. When the work calls for more. Comfortable with 8 GB+ VRAM.
Released November 15, 2026. GGUF, Q8_0. Runs with llama.cpp or any GGUF-compatible runtime.
| File | Quantization | Size | SHA-256 |
|---|---|---|---|
LMIM.Genesys-e3_w-soul-little-one-4B.Q8.gguf |
Q8_0 | 4.48 GB | c90aa7461b85935d9f8a3228ae1a87b1a250e69c80bdaf0c61d73fa3957eba09 |
sha256sum LMIM.Genesys-e3_w-soul-little-one-4B.Q8.gguf
# expected: c90aa7461b85935d9f8a3228ae1a87b1a250e69c80bdaf0c61d73fa3957eba09
On macOS: shasum -a 256 LMIM.Genesys-e3_w-soul-little-one-4B.Q8.gguf. The same hash is in SHA256SUMS in this repo, so sha256sum -c SHA256SUMS checks it in one line.
llama-cli -m LMIM.Genesys-e3_w-soul-little-one-4B.Q8.gguf \
-ngl 99 \
--jinja \
--conversation \
--temp 0.7 \
--top-p 0.9 \
--min-p 0.05 \
--repeat-penalty 1.12 \
--repeat-last-n 256 \
--reasoning off \
--reasoning-budget 1
Qwen/Qwen3.5-4B → Genesys e · 4B (v2, August 2026) → Genesys e3 · little one (November 2026, light fine-tune)
The model isn't the product. The continuity is.
iamonthemission/LMIM.Genesys-e_w-soul-4b