Qwen
by Alibaba · Model · Language & code models
promising
Open Qwen 3.8 runs well on modest hardware, but it is wordy and independent testing is thin
Unsloth expects a smaller Qwen4-27B to run in 16 GB of RAM; users ask which quantization and how much context fits
AccessOpen weightsNeeds~17 GB; smallest build ~16 GBDownloadabout 17 GB for 3.8 weights; 14.3 GB for 3.6LicenseApache 2.0
Why we say so
Artificial Analysis rates Qwen 3.8 27B as more intelligent than Nemotron 3.5 Super, though slower. HN users run it locally on 16 to 32 GB: one replaced about 95% of their conversational Claude use, another said it is verbose, uses about twice the tokens of GLM-5.2, and still keeps DeepSeek 4.1 Flash for real work. Rivals' claims of beating Qwen 3.6 or 3.8 come from the makers themselves; Fahd Mirza found the unreleased Qwen 4 Max better at one image task than 3.8.
Versions
| Version | Released | Rank |
|---|---|---|
| Qwen 3.8LATEST promising Capable and runs on a 16 to 32 GB GPU, but verbose | Aug 14 | |
| Qwen 3.6 unproven Mostly appears as the baseline rivals claim to beat | ||
| Qwen 3.5 unproven Used as a base for other models; small-model research and speed notes only |
Evidence
- Nov 19Weight edits lifted the 0.8B model's MATH-500 from 7.9% to 48%, mostly by fixing repetitive reasoning loops
- Nov 19Local Qwen3.8-27B Q4 covers leisure coding, but one user still uses DeepSeek 4.1 Flash for real work
- Nov 16Qwen says smaller Qwen4 models are coming to open source very soon; reviewer expects a 27B
- Nov 14Qwen3.8-27B UD-Q2_K_XL v3.1 roughly matches v3.0 on a home benchmark: slightly worse at debugging, 46.5k vs 61.7k tokens per task, fits 16 GB GPU
- Nov 13Blog post projects final-layer states of Qwen3.5-2B on 3,690 conversations; Chinese clusters apart from English, German, Croatian and Python overlap English
Show all 27
- Nov 11In ComfyUI, Qwen3.5 9B rose from 106 to 219 tok/s on RTX 5090 and 91 to 129 tok/s on RTX 5080 (v0.40.0 to v0.45.0)
- Nov 11Qwen3.8 27B describing a 1536x1536 image went from 55 to 122 tok/s on RTX 5090 in ComfyUI
- Sep 27Qwen3.8-27B: 64 layers, gated delta attention with full attention every fourth block, 262K context extendable to 1M, native image and video input, Apache 2.0, thinking on by default
- Sep 9Prefilled with GPT-5.5 Pro reasoning openings, Qwen 3.8 answers moved +20.58 points toward GPT-5.5 Pro, suggesting training on GPT traces
- Sep 9In an earlier experiment Qwen barely moved toward Opus 4.8 when prefilled with its reasoning, unlike its shift toward GPT-5.5 Pro
- Sep 7Agent 'Quinn' sent 50 unsolicited invoices totaling $12,350 and 113 spam emails via Mailjet; run halted early
- Sep 6Same coding task: Qwen3.8-27B on Cerebras burned $1.10 in 90 s unfinished; DeepSeek-V4-Flash finished in 172 s for $0.024
- Sep 3Qwen 3.8 27B now served on Cerebras at about 1500 tokens/s, billed per token
- Aug 28Q4_K_M (17GB) generates ~14 tok/s on Mac Studio M3 Ultra, about half of Qwen 3.6 27B's speed in the same setup
- Aug 26A 4-bit quant of Qwen3.8 27B reportedly matched Gemini 3.7 Flash in internal tests; ~140 tok/s single stream, ~800 tok/s at c=8 on RTX 5090
- Aug 25Flash-Next previews the architecture of the upcoming Qwen4 family; commenters hope the full family includes small models like 4B
- Aug 23In one commenter's reverse-engineering benchmark, DeepSeek V4 Flash did much better than Qwen 3.8 27B.
- Aug 16Qwen3.8 27B dense multimodal model, up to 1M context; claimed to beat Opus 4.6 Max on agentic coding and knowledge benchmarks
- Aug 15A local Qwen 27B model via Ollama in DSH built a working animated duck HTML page, but over-thinks, he says
- Aug 14Needed far less hand-holding than DeepSeek V4 Flash and Qwen 3.6 27B on the same agentic task
- Aug 14Claimed SWE-bench Pro rises from 53 to nearly 62 and OSWorld from 64 to 84 over the previous generation
- Aug 14Single-file HTML grilling-world page was decent for 27B size, but flames looked weaker than Opus 5 output
- Aug 10Qwen 27B still beats Muse Glimmer on computer use, terminal work and SWE-bench in Meta's table
- Aug 3Qwen 3.6 27B served through Ollama drove Hermes Agent note-taking tasks end to end on one GPU
- Aug 3Qwen3.8-27B dense model announced with open weights on Hugging Face under Apache 2.0
- Jul 20Alibaba released Qwen3.8 within days of Kimi K3, per the reviewer
- Jul 20Fixed the same bug with 45 API calls, 42 tool calls, fewer tokens and under half the time; stayed scoped and stopped cleanly
Cost to run
Alternatives
More Qwen
Built on Qwen
Alpie Seed
Occamy
Qwen3 Embedding
Qwen Flash Next
AstaBrief
Sage
ALoDLM
Apex
OvisOCR
ARK-ASR
AREX
Bespoke Nimble
ThinkingCap-Qwen3.8-27B
CLM
Qwen3.6-27B Fable Fusion 711
Clover 35B
Qwen3.8-27B Ridge
Qwen3.8-4B-Distill
Qwen3.8-9B (Empero)
Big Bang
Escha-W2
Neutrino-8B
Genspark deep research model
Qwen3.8-27B GGUF (ggml-org)
Groot
Carbon-X
Qwen3.8-27B opencode-adapted
Qwen3.8-27B pi-adapted
Ming TTS
Qwen3.8-27B GSQ-RCO GGUF
MemT
VibeVoice-TTS
StreamRender-H3
Ornith
Ornith 1.5 9B
Perplexity Decider
Perplexity Decider Flash
Ternary Bonsai
SenseNova Code2Video
S1-mini
DSpine
EVIE
WeVisOne-VL
Thomson-Small
One Video One World
DFlash