Qwen3.8 Flash Next
Alibaba · Qwen/Qwen3.8-Flash-Next
Q4 file 93.7 GB (UD-IQ4_XS) · Weights license: Publisher terms: Qwen Community 1.0
Quality
Quality · Overall
LiveBench 2026-06-25 · 0–100 · higher is better · this model outlined, #9 of 23
- Color = publisher
- Memory under review
Source: LiveBench 2026-06-25 · collected 2026-10-18 · open models without a score are not drawn (no score ≠ 0)
No result in LMArena 2026-10-22 Not in LMArena 2026-10-22
LiveBench categories
This model's mean per category · 0–100
Source: LiveBench 2026-06-25 · collected 2026-10-18 · open models without a score are not drawn (no score ≠ 0)
Reported by the publisher
- Agentic coding DeepSWE 1.1
- 58.7Language
- Agentic coding SWE-bench Pro
- 62.5Language
- Multilingual software engineering SWE-bench Multilingual
- 81.0Language
- Repo-level code generation NL2Repo-Bench
- 48.1Language
- Long-horizon office work CoWorkBench
- 73.9Language
- Professional job tasks JobBench
- 55.7Language
- Real-world tool use Toolathlon Verified (Pass@1)
- 73.5Language
- Instruction following IFBench
- 81.3Language
Every reported result
- Scientific reasoning GPQA Diamond
- 91.7Language
- Multidisciplinary reasoning HLE
- 35.9Language
- Competitive coding LiveCodeBench v6
- 91.9Language
- Application recreation RecreationBench
- 49.9Vision Language
- Mobile use AndroidWorld
- 84.5Vision Language
- Visual web development Vision2Web
- 64.0Vision Language
- Embodied intelligence ERQA
- 72.3Vision Language
- Long video understanding LVBench
- 76.6Vision Language
- Real-world perception RealWorldQA
- 88.5Vision Language
Values as printed on the card, with the publisher's own settings (card section: Benchmark Results › Language; Benchmark Results › Vision Language); column “Qwen3.8-Flash-Next”. Not comparable across publishers and not used in any ranking. Model card
Run it
Run it
File UD-IQ4_XS · 93.7 GB · 3 parts · unsloth/Qwen3.8-Flash-Next-GGUF · What is Q4?
llama-server -hf unsloth/Qwen3.8-Flash-Next-GGUF:UD-IQ4_XS- Install llama.cpp
- Run the command above
- Ask at localhost:8080:
Explain what a context window is, in two sentences.
Quality × memory
Not placed on quality × memory
There is no validated Q4 planning estimate for this checkpoint (its parameter basis does not allow one).
Runs on
No memory requirement to match
Without a parameter count with a known basis or a documented memory path, no hardware is grouped. Nothing is guessed.
Qwen3.8 Flash Next is a text model from Alibaba; this catalog has no verified parameter count for it, so no memory is estimated. It was released on August 26, 2026 and reads up to 262,144 tokens of context. We collected 12 downloadable GGUF and MLX files from 2 publisher sources; the most downloaded GGUF source is unsloth's. It scores 76.2 overall in LiveBench 2026-06-25.
Compare Qwen3.8 Flash Next using its publisher card and the LiveBench 2026-06-25 snapshot. This is quality evidence, not local speed.
This profile has no simplified memory estimate because the available size evidence is incomplete or includes separate components. Check the exact checkpoint, format and runtime before planning hardware.
Versions to download
Files you can download for this checkpoint. The section opens on the most downloaded collected GGUF source, unsloth, at UD-IQ4_XS. On a Mac it opens on the most downloaded MLX source. Sizes are the Hub's published bytes. What is Q4?
unsloth
unsloth/Qwen3.8-Flash-Next-GGUF · UD-IQ4_XS/Qwen3.8-Flash-Next-UD-IQ4_XS-00001-of-00003.gguf · 93.7 GB
Run this file
llama-server -hf unsloth/Qwen3.8-Flash-Next-GGUF:UD-IQ4_XSlms get https://huggingface.co/unsloth/Qwen3.8-Flash-Next-GGUF- Ollama: Ollama's documented hf.co form covers single-file GGUFs; this file has 3 parts
- mlx-lm: mlx-lm runs MLX repositories; pick the MLX version on a Mac
Quality scores on this page belong to the original checkpoint; these files are not scored or ranked. File sizes read 2026-10-23 · download counts read 2026-10-17 · commands: Hugging Face Hub docs: Use Ollama with any GGUF model · llama.cpp server README (-hf <user>/<model>[:quant]) · LM Studio docs: lms get with a Hugging Face URL · mlx-lm README (mlx_lm.generate --model)
Fine-tunes, adapters and other versions
- Quantized versions
- 500
- Fine-tunes
- 94
- Adapters · merges
- 2 · 7
Browse them on the Hugging Face model tree · read 2026-10-23
Uncensored and abliterated versions
Some community uploads remove the model's safety refusals. Quality may drop, the original publisher does not support them, claims about them are the uploader's, and the original license still applies. This site never recommends or ranks them.
59 derivative repositories have “uncensored” or “abliterated” in their name (read 2026-10-23).
Measured speed
No published measurement for this checkpoint
Tokens per second are shown only when a source publishes them with hardware, build and setup. They are never estimated.
How to run
No validated memory estimate for this model and runtime.
Q4 planning estimate only. Check checkpoint format, runtime support, KV cache, context, batch size and offload before choosing hardware.
Evidence level: Generic estimate only; no execution reproduced by this site.
- Publisher sampling defaults
- temperature 1 · top_p 0.95 · top_k 20generation_config.json · read 2026-10-23source
- Chat template markers
- tool calls · thinkingmarkers found in the template, not a test of the feature · read 2026-10-23source
Before downloading
- Check the exact checkpoint and license linked here.
- Confirm a compatible runtime and precision or quantization for your platform.
- Allow for context, batch size and any extra media components beyond the weights.
Technical details
What the source and the catalog store for this checkpoint. Missing fields are left out, never zero.
- Checkpoint / version
- Qwen/Qwen3.8-Flash-Next
- Task and modality
- language and reasoning · text + image + video
- Count and basis
- Not verified
- Context
- 262,144 tokens published for this model
- Architecture
- MoEQwen/Qwen3.8-Flash-Next model card (Hugging Face) · read 2026-10-09source
- Released
- QwenLM/Qwen3.8-Flash-Next README (GitHub) · read 2026-10-13source
- Runtime and precision
- No local execution verified.
- Commercial use
- Allowed with conditions Any model-as-a-service or work-assistant business needs a Qwen license; large products show the model name · Qwen Community License 1.0 (Qwen/Qwen3.8-Flash-Next LICENSE) · read 2026-10-18license text
Every stored field
- Revision checked
- de4b8e4d43b917e7706784d8bb445c9af86a3540
- Architecture
- Qwen4ExpForConditionalGeneration
- Library
- transformers
- Hub task tag
- image-text-to-text
- Distributions in this group
- mlx-community/Qwen3.8-Flash-Next-4bit-mtp
- Formats
- safetensors
- Repository created
- 2026-08-24creation is not the model's release
- Repository updated
- 2026-08-27a commit is not a release date
- Hub downloads (30 days) · likes
- 3,093,369 · 6,473this repository only · read 2026-10-25
- Read from the Hub page
- task, transformers class, task tags, license name, file formatsHub overview and file listing · read 2026-10-18
- Catalog state
- Hub inventory current
125B backbone + 51B n-gram memory + 4B MTP. Local-memory planning needs an architecture-specific review; no simplified estimate is shown.
Sources
Identity and published specifications come from the checkpoint page. Benchmark scores belong to the named release and do not measure speed on your computer.
- Publisher checkpoint and model card
- Weight license and use terms
- Quality benchmark source · LiveBench 2026-06-25 overall
Catalog reviewed: 2026-09-30. Benchmark release: 2026-06-25; snapshot collected: 2026-10-18.