---
license: apache-2.0
language:
  - en
tags:
  - text-to-sql
  - data-chat
  - gguf
  - qwen
  - qwen3.5
base_model: Qwen/Qwen3.5-122B-A10B
---

# tribeblend-etl-qwen35-122b-a10b

Qwen3.5 122B-A10B MoE post-trained chat model adapted for users with very high-memory local systems.

Direct **base-model** GGUF (Q4_K_M) of `Qwen/Qwen3.5-122B-A10B`, published for
TribeBlend's local Data Chat runtime. TribeBlend grounds answers with Knowledge
Graph context at prompt time and a model-aware agent harness, so the base
instruction/reasoning model ships as-is (no fine-tuning).

- **Base model**: Qwen/Qwen3.5-122B-A10B
- **Provider / family**: qwen / qwen3.5
- **Local runtime arch**: qwen3.5
- **Recommended profile**: expert
- **Quantization**: Q4_K_M
- **Native context window**: 262144

## Usage

Designed for [TribeBlend](https://github.com/TribeBlend/tribeblend) Data Chat,
loaded via llama-cpp-2.

## License

Inherits the upstream base-model license (apache-2.0); verify upstream
terms before redistribution.


## Split GGUF files

This Q4_K_M GGUF is split into multiple shards because the Hugging Face repository enforces a 50 GB maximum individual file size. Download all shard files in this repo and load the first shard (`tribeblend-etl-qwen35-122b-a10b-q4_k_m-00001-of-00002.gguf`) with a GGUF runtime that supports split GGUF files.