id: qwen3-coder-30b-a3b-instruct-q6_k
name: Qwen3 Coder 30B A3B Instruct GGUF Q6_K
type: gguf-model
task:
  - 1c
  - code
  - agentic-coding
  - repository-analysis
  - tool-use
language:
  - ru
  - en
source: huggingface
upstream_id: lmstudio-community/Qwen3-Coder-30B-A3B-Instruct-GGUF
upstream_url: https://huggingface.co/lmstudio-community/Qwen3-Coder-30B-A3B-Instruct-GGUF
license: apache-2.0
status: candidate
created_at: 2026-06-19
storage_path: /models/gguf/1c/qwen3-coder-30b-a3b-instruct-q6_k
filename: Qwen3-Coder-30B-A3B-Instruct-Q6_K.gguf
file_size_bytes: 25104724288
format: gguf
quantization: Q6_K
context_length: 262144
deployment_context_length: 8192
vram_required_gb: null
parameters: 30.5B total / 3.3B active
base_model: Qwen/Qwen3-Coder-30B-A3B-Instruct
adapter_for: null
datasets: []
eval_suites:
  - plugins/1c/evals/smoke.yaml
deployment:
  target: docker-gpu.cin.su
  runtime: llama.cpp
  served_model_name: qwen3-coder-1c-q6
  compose: core/deploy/docker-gpu/llama-cpp/compose.yaml
notes: "High-quality Q6_K quant for local 1C/code experiments on a single RTX 4090. Expected to require partial CPU/RAM offload or a short context because the GGUF file is larger than 24 GiB VRAM. Test first at 8K context, then 16K if stable."