v2.2 · 2026-06-12

Qwen3-Coder-Next 80B A3B

qwen3-coder-next-80b-a3b

CODE

Qwen3-Coder-Next 80B A3B is a 80B code model profile for local AI planning. This page records provider, license, context, quantization, hardware-fit estimates, setup hints, source links, and caveats so readers do not choose a model by name alone.

PROVIDER

Alibaba Qwen

FAMILY

Qwen3-Coder-Next

MODEL TYPE

code

PARAMETERS

80B (3B active)

MODALITIES

text, code

ARCHITECTURE

Hybrid attention plus Gated DeltaNet plus sparse MoE causal language model; 512 experts, 10 activated plus 1 shared expert; non-thinking instruct model

CONTEXT WINDOW

262,144 tokens

TRAINING TOKENS

—

RELEASE DATE

—

LICENSE & LINKS

Apache-2.0license text

MODEL CARD

huggingface.co/Qwen/Qwen3-Coder-Next

HUGGING FACE

Qwen/Qwen3-Coder-Next

GGUF REPOSITORIES

Qwen/Qwen3-Coder-Next-GGUF official

Official GGUF repository referenced by the June 2026 model research.

unsloth/Qwen3-Coder-Next-GGUF community conversion

Community GGUF repository referenced by the June 2026 model research.

SETUP HINTS

OLLAMA

ollama run qwen3-coder-next

LM STUDIO

Available according to the supplied June 2026 research; select a cited GGUF/MLX build and verify current app metadata before relying on hands-on setup guidance.

LLAMA.CPP

Use the cited GGUF repository (Qwen/Qwen3-Coder-Next-GGUF) with current llama.cpp-compatible tooling.

QUANTIZATION FILE SIZES (GGUF)

Q4_K_M

48.5 GB

Q5_K_M

56.8 GB

Q8_0

84.8 GB

Quantization note: Q4_K_M for first local attempts unless a cited runtime recommends a different default.

RAM / VRAM ESTIMATES

MIN RAM

42 GB

COMFORTABLE RAM

64 GB

MIN VRAM

42 GB

COMFORTABLE VRAM

64 GB

These are conservative local-inference estimates, not official hardware requirements.

HARDWARE FIT

CPU only (no GPU)

8 GB RAM

16 GB RAM

32 GB RAM

8 GB VRAM

12 GB VRAM

24 GB VRAM

Apple Silicon (unified memory)

Hardware fit values are conservative local-inference estimates based on GGUF size plus runtime and KV-cache overhead. Actual requirements depend on context length, quantization, runtime, GPU offload, and other running apps.

BEST FOR

·Coding agents, terminal automation, repo edits, and tool calling on 64GB plus workstations or Macs.

AVOID IF

·You have 32GB RAM, expect laptop speed, or need frictionless beginner setup immediately without tuning.

CAVEATS

·Although only 3B parameters are active, memory requirements are driven by total model storage, quantization, runtime overhead, and context length.

·Full 256K context may require reducing context length on local systems to avoid out-of-memory errors.

·Best used with current SGLang, vLLM, llama.cpp, Ollama, or LM Studio builds that understand Qwen3-Coder tool formatting.

·Non-thinking model; it is intended for coding-agent behavior, not visible reasoning traces.

·Commercial-use status from the supplied research: Permitted under Apache-2.0 terms

·Beginner summary from supplied research: Qwen3 Coder Next targets local coding agents: an 80B MoE with only 3B active parameters, strong tool use, and practical workstation deployment when 30B feels underpowered but 480B is impossible for most local builders today.

SOURCE URLS