Total runs:
350.9K
Run Growth:
273.7K
Growth Rate:
78.40%
Sponsored by
APIMart
AtomicChat / Qwen3.8-Flash-Next-GGUF
Model's Last Updated: August 27 2026
text-generation
Introduction of Qwen3.8-Flash-Next-GGUF
Model Details of Qwen3.8-Flash-Next-GGUF
How to Run Qwen3.8-Flash-Next Locally
Built from Qwen's original weights with our own importance matrix. The calibration corpora behind our builds are public.
- Qwen3.8-Flash-Next is the first open-weight release of the architecture behind Qwen4.
- These GGUFs are self-quantized from Qwen's original weights with our own importance matrix, published alongside the quants.
- The quants are still uploading and need a llama.cpp build with Qwen3.8-Flash-Next support; Atomic Chat runs it as support ships.
Qwen3.8-Flash-Next architecture (Qwen).
Highlights
- 125B total with 6B active sparse MoE (512 experts, 10 routed + 1 shared), plus a 51B n-gram embedding and a 4B MTP layer. An experimental preview of the architecture behind Qwen4.
- Hybrid attention with QSA : Gated DeltaNet paired with Qwen Sparse Attention, which operates at the micro-block level rather than per token to cut long-context latency for agentic workloads.
- Gated Residual : a data-dependent read gate plus a per-branch scalar write gate over widened residual streams, for finer expressiveness at low inference overhead.
- N-gram Embedding : 20M bigram/trigram embeddings indexed at layer 2, a compute-light axis for parameter scaling that offloads well on memory-constrained accelerators.
- 262,144-token context , extensible up to 1,000,000 tokens with RoPE scaling.
- Natively multimodal (causal language model with a vision encoder, image-text-to-text). These GGUF quants cover the text path.
- Frontier coding and agentic scores (Qwen-reported): LiveCodeBench v6 91.9, GPQA Diamond 91.7, SWE-bench Multilingual 81.0, CoWorkBench 73.9.
- Full imatrix quantization with our public calibration corpora .
These GGUFs are self-quantized from the original weights , not a repack. The importance matrix keeps low-bit quants closer to the full-precision model.
Always pass
--jinjaso the Qwen3.8-Flash-Next chat template is applied. Without it the model can emit malformed turns.
Model Overview
| Property | Value |
|---|---|
| Base model |
Qwen/Qwen3.8-Flash-Next
|
| Type | Causal language model with a vision encoder (image-text-to-text) |
| Total / active parameters | 125B total / 6B active, plus 51B n-gram embedding and a 4B MTP layer |
| Layers | 48. Hidden layout: 12 x (3 x (Gated DeltaNet then MoE) then 1 x (Qwen Sparse Attention then MoE)) |
| Experts | 512 experts, 10 routed + 1 shared activated |
| Attention | Hybrid: Gated DeltaNet (linear) and Qwen Sparse Attention (micro-block sparse); Gated Residual over widened residual streams |
| Context length | 262,144 native, extensible up to 1,000,000 |
| This repo | GGUF quants (imatrix), text path. The importance matrix we built is published here too. |
Scores are Qwen's published results for the base
Qwen/Qwen3.8-Flash-Next
. Quantization preserves the large majority of this;
Q4_K_M
and up sit within a point or two of full precision.
Choosing a quant
| Quant | Size | Notes |
|---|---|---|
IQ2_M
|
— | Smallest usable. Aggressive low-bit for memory-constrained boxes. |
IQ3_M
|
— | Beats Q3 at similar size thanks to imatrix. Best low-RAM pick. |
Q4_K_M
|
— | Recommended default. Best balance of size, speed and quality. |
UD-Q4_K_XL
|
— | Dynamic. Embeddings and output kept at Q8_0 for higher quality at a Q4 footprint. |
Q6_K
|
— | Near lossless. |
Q8_0
|
— | Effectively lossless, reference quality. |
Sizes fill in once the quants finish uploading. Pick the largest file that fits your (V)RAM with room for context.
Get started
Qwen3.8-Flash-Next is a brand-new Qwen4-preview architecture (Gated DeltaNet, Qwen Sparse Attention, n-gram embedding). The quants in this repo are still uploading, and running them needs a
llama.cppbuild that has landed Qwen3.8-Flash-Next support. Until then, Atomic Chat is the easiest way to run it as support ships.
Run Qwen3.8-Flash-Next locally with:
-
Atomic Chat
:
the easiest path. Open the app, search
AtomicChat/Qwen3.8-Flash-Next-GGUF, pick a quant, hit Use this model . -
llama.cpp:
llama-server -hf AtomicChat/Qwen3.8-Flash-Next-GGUF:Q4_K_M --jinja -c 8192 -
Ollama:
ollama run hf.co/AtomicChat/Qwen3.8-Flash-Next-GGUF:Q4_K_M - LM Studio / Jan: search the repo id, download any quant.
Best practices
| Parameter | Value |
|---|---|
| temperature | 1.0 |
| top_p | 0.95 |
| top_k | 20 |
| min_p | 0.0 |
Qwen's recommended thinking-mode settings. For non-thinking (instruct) use
temperature=0.7
,
top_p=0.80
,
top_k=20
,
presence_penalty=1.5
. Allocate generous output length for agentic tasks.
Run in llama.cpp
git clone https://github.com/ggerganov/llama.cpp
cmake llama.cpp -B llama.cpp/build -DBUILD_SHARED_LIBS=OFF -DGGML_CUDA=ON
cmake --build llama.cpp/build --config Release -j --target llama-cli llama-server
./llama.cpp/build/bin/llama-server \
-hf AtomicChat/Qwen3.8-Flash-Next-GGUF:UD-Q4_K_XL \
--jinja -ngl 99 -c 8192 -fa on
How these were made
-
Download
Qwen/Qwen3.8-Flash-Next(original weights). - Convert to GGUF with a llama.cpp build that supports the Qwen3.8-Flash-Next architecture (Gated DeltaNet, Qwen Sparse Attention, n-gram embedding).
- Build an importance matrix over our public calibration corpora .
-
Quantize the ladder with
--imatrix;UD-Q4_K_XLadditionally pins the token-embedding and output tensors toQ8_0.
License
Released by Qwen under the Qwen Community License 1.0. Quantized by Atomic Chat.
Runs of AtomicChat Qwen3.8-Flash-Next-GGUF on huggingface.co
226.1K
Total runs
0
24-hour runs
55.6K
3-day runs
149.3K
7-day runs
197.8K
30-day runs
More Information About Qwen3.8-Flash-Next-GGUF huggingface.co Model
Url of Qwen3.8-Flash-Next-GGUF
Qwen3.8-Flash-Next-GGUF huggingface.co Url
Provider of Qwen3.8-Flash-Next-GGUF huggingface.co
Other API from AtomicChat
Updated:August 09 2026
Total runs:
118.8K
Run Growth:
133.7K
Growth Rate:
64.83%
Total runs:
91.0K
Run Growth:
86.5K
Growth Rate:
90.45%
Total runs:
73.4K
Run Growth:
43.9K
Growth Rate:
59.85%
Total runs:
73.1K
Run Growth:
43.0K
Growth Rate:
58.76%
Total runs:
72.9K
Run Growth:
42.4K
Growth Rate:
58.17%
Total runs:
61.7K
Run Growth:
60.4K
Growth Rate:
95.16%
Total runs:
6.6K
Run Growth:
1.4K
Growth Rate:
20.56%
Total runs:
5.1K
Run Growth:
-251
Growth Rate:
-4.89%
Total runs:
4.3K
Run Growth:
-3.4K
Growth Rate:
-78.42%
Total runs:
4.2K
Run Growth:
-13.8K
Growth Rate:
-331.35%
Total runs:
4.1K
Run Growth:
-8.3K
Growth Rate:
-201.33%
Total runs:
3.7K
Run Growth:
-2.2K
Growth Rate:
-60.40%
Total runs:
3.7K
Run Growth:
-11.3K
Growth Rate:
-305.72%
Total runs:
3.2K
Run Growth:
-17.5K
Growth Rate:
-552.69%
Total runs:
3.1K
Run Growth:
566
Growth Rate:
19.19%
Total runs:
3.0K
Run Growth:
-2.3K
Growth Rate:
-75.56%
Total runs:
2.9K
Run Growth:
-38
Growth Rate:
-1.32%
Total runs:
2.5K
Run Growth:
-1.5K
Growth Rate:
-60.92%
Total runs:
2.3K
Run Growth:
-8.7K
Growth Rate:
-376.96%
Total runs:
2.2K
Run Growth:
516
Growth Rate:
23.63%
Total runs:
2.0K
Run Growth:
-13.9K
Growth Rate:
-680.84%
Total runs:
1.8K
Run Growth:
1.7K
Growth Rate:
100.00%
Total runs:
1.8K
Run Growth:
692
Growth Rate:
38.44%
Total runs:
1.8K
Run Growth:
-14.8K
Growth Rate:
-832.36%
Total runs:
1.7K
Run Growth:
657
Growth Rate:
37.65%
Total runs:
1.6K
Run Growth:
-484
Growth Rate:
-30.02%
Total runs:
1.6K
Run Growth:
1.6K
Growth Rate:
100.00%
Total runs:
1.5K
Run Growth:
687
Growth Rate:
45.02%
Total runs:
1.2K
Run Growth:
1.2K
Growth Rate:
100.00%
Total runs:
1.1K
Run Growth:
-109
Growth Rate:
-9.57%
Total runs:
1.1K
Run Growth:
-2.1K
Growth Rate:
-184.69%
Total runs:
1.1K
Run Growth:
-931
Growth Rate:
-82.76%
Total runs:
1.1K
Run Growth:
1.1K
Growth Rate:
100.00%
Total runs:
1.0K
Run Growth:
1.0K
Growth Rate:
100.00%
Total runs:
1.0K
Run Growth:
1.0K
Growth Rate:
100.00%
Total runs:
727
Run Growth:
-71
Growth Rate:
-9.77%
Total runs:
658
Run Growth:
-7.5K
Growth Rate:
-1235.47%
Total runs:
637
Run Growth:
-3.4K
Growth Rate:
-566.16%
Total runs:
635
Run Growth:
635
Growth Rate:
100.00%
Total runs:
632
Run Growth:
-3.5K
Growth Rate:
-559.34%
Total runs:
603
Run Growth:
-61
Growth Rate:
-10.12%
Total runs:
602
Run Growth:
602
Growth Rate:
100.00%
Total runs:
590
Run Growth:
590
Growth Rate:
100.00%
Total runs:
565
Run Growth:
101
Growth Rate:
17.88%
Total runs:
525
Run Growth:
-176
Growth Rate:
-33.52%
Total runs:
521
Run Growth:
187
Growth Rate:
35.89%
Total runs:
512
Run Growth:
-1.2K
Growth Rate:
-237.89%
Total runs:
456
Run Growth:
-576
Growth Rate:
-126.32%
Total runs:
429
Run Growth:
-426
Growth Rate:
-99.30%
Total runs:
416
Run Growth:
416
Growth Rate:
100.00%
Total runs:
413
Run Growth:
-1.2K
Growth Rate:
-300.48%
Total runs:
384
Run Growth:
316
Growth Rate:
82.29%
Total runs:
382
Run Growth:
-209
Growth Rate:
-54.71%
Total runs:
381
Run Growth:
-1.4K
Growth Rate:
-369.82%
Total runs:
313
Run Growth:
248
Growth Rate:
79.23%
Total runs:
303
Run Growth:
303
Growth Rate:
100.00%
Total runs:
295
Run Growth:
295
Growth Rate:
100.00%
Total runs:
293
Run Growth:
-639
Growth Rate:
-218.09%
Total runs:
292
Run Growth:
292
Growth Rate:
100.00%
Total runs:
285
Run Growth:
-147
Growth Rate:
-51.58%
Total runs:
267
Run Growth:
-344
Growth Rate:
-128.84%
Total runs:
250
Run Growth:
-1.5K
Growth Rate:
-601.60%
Total runs:
247
Run Growth:
-417
Growth Rate:
-168.83%
Total runs:
235
Run Growth:
235
Growth Rate:
100.00%
Total runs:
224
Run Growth:
-853
Growth Rate:
-380.80%
Total runs:
220
Run Growth:
-1.5K
Growth Rate:
-684.55%
Total runs:
219
Run Growth:
-208
Growth Rate:
-94.98%
Total runs:
218
Run Growth:
-61
Growth Rate:
-27.98%
Total runs:
208
Run Growth:
208
Growth Rate:
100.00%
Total runs:
192
Run Growth:
-123
Growth Rate:
-64.06%
Total runs:
191
Run Growth:
-6.4K
Growth Rate:
-3354.97%
Total runs:
159
Run Growth:
-623
Growth Rate:
-391.82%
Total runs:
152
Run Growth:
-901
Growth Rate:
-592.76%
Total runs:
135
Run Growth:
-254
Growth Rate:
-188.15%
Total runs:
135
Run Growth:
-666
Growth Rate:
-493.33%
Total runs:
135
Run Growth:
-102
Growth Rate:
-75.56%
Total runs:
134
Run Growth:
-341
Growth Rate:
-254.48%
Total runs:
120
Run Growth:
-428
Growth Rate:
-356.67%
Total runs:
117
Run Growth:
117
Growth Rate:
100.00%
Total runs:
117
Run Growth:
-885
Growth Rate:
-756.41%
Total runs:
117
Run Growth:
104
Growth Rate:
88.89%
Total runs:
106
Run Growth:
-757
Growth Rate:
-714.15%
Total runs:
105
Run Growth:
-803
Growth Rate:
-764.76%
Total runs:
104
Run Growth:
-1.0K
Growth Rate:
-992.31%
Total runs:
104
Run Growth:
-201
Growth Rate:
-193.27%
Total runs:
101
Run Growth:
-911
Growth Rate:
-901.98%
Total runs:
100
Run Growth:
-338
Growth Rate:
-338.00%
Total runs:
94
Run Growth:
82
Growth Rate:
87.23%
Total runs:
93
Run Growth:
-1.2K
Growth Rate:
-1288.17%
Total runs:
82
Run Growth:
-366
Growth Rate:
-446.34%
Total runs:
78
Run Growth:
-974
Growth Rate:
-1248.72%
Total runs:
65
Run Growth:
-14
Growth Rate:
-21.54%
Total runs:
64
Run Growth:
-29
Growth Rate:
-45.31%
Total runs:
57
Run Growth:
-33
Growth Rate:
-57.89%
Total runs:
47
Run Growth:
-1
Growth Rate:
-2.13%
Total runs:
45
Run Growth:
-15
Growth Rate:
-33.33%