Downloads · 30 days
10
22% of all-time downloads
aweussom/Qwen3-30B-A3B-Thinking-2507-int4-ov
Qwen3-30B-A3B-Thinking-2507-int4-ov is a machine learning model from aweussom. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
INT4 OpenVINO export of Qwen/Qwen3-30B-A3B-Thinking-2507, built to pair with Intel's own OpenVINO/Qwen3-30B-A3B-Instruct-2507-int4-ov: same architecture, same quantization recipe, so the two variants are directly
Downloads · 30 days
10
22% of all-time downloads
All-time downloads
46
Public
Repo size
16.3 GB
Likes
0
Public
Click a slice to open those files.
.bin16.3 GB · 100%
From the Hugging Face model README
INT4 OpenVINO export of Qwen/Qwen3-30B-A3B-Thinking-2507, built to pair with Intel's own OpenVINO/Qwen3-30B-A3B-Instruct-2507-int4-ov: same architecture, same quantization recipe, so the two variants are directly comparable in an A/B — reasoning on/off is the only variable.
The 2507 generation split Qwen3's hybrid thinking mode into two committed
models: Instruct (never thinks) and Thinking (always thinks). Intel published a
pre-converted Instruct but, as of 2026-08, no Thinking counterpart. This fills
that gap with the identical recipe (INT4 asymmetric, group size 128 — verified
against the Instruct export's rt_info).
optimum-cli export openvino \
--model Qwen/Qwen3-30B-A3B-Thinking-2507 \
--weight-format int4 \
--trust-remote-code \
Qwen3-30B-A3B-Thinking-2507-int4-ov
Toolchain: OpenVINO 2026.1.0-21367, optimum-intel 1.27.0, transformers 4.57.6, NNCF weight compression (98% of layers int4_asym gs128, embeddings/head int8_asym per-channel). 15.2 GB on disk.
openvino_model.bin
(sha256 f6afe8bb73031387a1831dce0d987210311e47d1dbc38b3a47d642ef62697d09)
is byte-identical to the independently published
livrrr/Qwen3-30B-A3B-Thinking-2507-int4-ov.
Two independent runs of the official export pipeline agree bit-for-bit —
NNCF data-free int4 compression is deterministic — so neither repo asks for
trust: verify the hash yourself.<think> tag, so responses contain reasoning text terminated by </think>,
then the answer. Parse accordingly.max_new_tokens can be consumed entirely by reasoning, returning a reply
with no post-</think> content. Give it room (2k+ for short answers, much
more for long-form tasks).OFFLOAD_RATIO streams expert weights from disk to fit tight VRAM budgets;
without XMX hardware that property is a silent no-op.python nollama.py --model-dir Qwen3-30B-A3B-Thinking-2507-int4-ov --device CPU --cache-size-gb 12
NoLlama is an OpenAI-compatible
LLM/VLM server for Intel hardware; --cache-size-gb sizes the KV pool
(~12 GB holds a ~95k-token prompt for this geometry).