Davd-b01/transductor-high-v3-gguf overview
Transductor TC High — GGUF Q4 K M Quantized GGUF of Davd b01/transductor high v3 SFT+SimPO full merge, LFM2.5 2.6B for local CUDA/Vulkan inference. Q4 K M veri…
Runs locally from ~1.56 GB disk (4 GB VRAM class GPUs with llama.cpp / guIDE).
Repository Files & Downloads
| File | Type | Quantization | Size | Link |
|---|---|---|---|---|
| high-sftplus-simpo-q4_k_m.gguf | GGUF | Q4_K_M | 1.56 GB | Download |
Model Details
| Model ID | Davd-b01/transductor-high-v3-gguf |
|---|---|
| Author | Davd-b01 |
| Pipeline | text-generation |
| License | — |
| Base model | Davd-b01/transductor-high-v3 |
| Last modified | 2026-09-07T01:53:35.000Z |
Model README
---
library_name: gguf
pipeline_tag: text-generation
base_model: Davd-b01/transductor-high-v3
tags:
- lfm2
- reasoning-traces
- thinking-levels
- gguf
---
Transductor TC High — GGUF (Q4_K_M)
Quantized GGUF of Davd-b01/transductor-high-v3 (SFT+SimPO full merge, LFM2.5-2.6B) for local CUDA/Vulkan inference. Q4_K_M verified quality-neutral vs Q8_0 on a 10-item harness check; smoke-tested think+answer output.
Usage: llama-server -m high-sftplus-simpo-q4_k_m.gguf -ngl 999 -fa on with the tier system prompt + four-slot TCS input; parse <tc_think> / <tc_answer>.
Run Davd-b01/transductor-high-v3-gguf with guIDE
Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.
Source: Hugging Face · Compare models