Last updated:
Qwen4-27B vs Qwen3.8-27B
Qwen3.8-27B is the open 27B Qwen model you can run today. Qwen4-27B is announced but not released, so most of this comparison is “not published yet”. The useful question is whether to wait.
Side by side
| Qwen3.8-27B | Qwen4-27B | Qwen3.8-Flash-Next (Qwen4 preview) | |
|---|---|---|---|
| Status | Released (Aug 2026) | Announced, not released | Released (Aug 2026) |
| Design | Dense, hybrid attention | Not published | Mixture of experts |
| Parameters | 27B | Not published | 125B total, about 6B active, plus 51B N-gram |
| Context | 262K, up to about 1M with YaRN | Not published | 262K, up to about 1M with YaRN |
| Input | Text, image, video | Not published | See model card |
| License | Apache 2.0 | Not published | qwen-community-1.0 |
| 4-bit, 32K VRAM | about 20 GB | about 20 GB if similar | Much more: all 125B parameters must be loaded |
Source: Qwen3.8-27B model card
Source: Qwen3.8-Flash-Next model card
Source: DataLearner: Qwen4-27B status
Should you wait for Qwen4-27B?
- Deploy Qwen3.8-27B now. If you need a working local model this month. Everything you set up, from GPU and runtime to prompts and API wrapper, carries over.
- Wait. If you are about to buy hardware. Qwen4-27B’s memory needs are unknown until release, and a switch to a mixture-of-experts design would change them.
More on Qwen4
Qwen4 status and release date · Qwen4-27B · VRAM calculator · GGUF downloads · Run locally · Try online
FAQ
Is Qwen4-27B better than Qwen3.8-27B?
Unknown. Alibaba has published no Qwen4 benchmarks.
Will Qwen4-27B use the same hardware?
Likely, if it stays a 27B dense-style model. Alibaba has not published its architecture, so confirm after release.
Is Qwen3.8-Flash-Next a smaller Qwen4?
No. It is a larger mixture-of-experts preview of the Qwen4 architecture, with 125B total parameters.
Independent guide, not affiliated with Alibaba or the Qwen team. Status checked on October 11, 2026. Qwen4 weights are not released yet; we will update this page when they are.