At a glance
- Maker: Alibaba
- Released: 2026
- Parameters: 27B
- Context window: 262k
- Size: Medium
- Memory: Handles very long documents
- Good at: Coding and careful answers
- Supports: tools, images, zero data retention
- License: Apache 2.0
- Upstream API price: $3.20 per million tokens
Announced in The Qwen3.8 family announcement.
What is Qwen3.8 27B?
Qwen3.8 27B is the newest mid-size model in Alibaba's Qwen3.8 generation, and it takes the opposite bet from its Flash sibling: rather than sparsity and speed, it is a dense 27-billion-parameter model whose full weight contributes to every token. The family's own summary of it is that it thinks before it answers — it is a reasoning model, and on UsingOpen it runs with a low default reasoning effort so that care does not turn into slowness.
Dense at this scale has two practical consequences. First, quality per parameter tends to be high: there are no sparse corners to cut, which is why mid-size dense models remain favourites for careful work. Second, the economics differ from the mixture-of-experts crowd — at $3.20 per million tokens on Alibaba's API it costs more than the Flash tier, which is the trade you make for a model with all of its parameters active on every answer. The 262k-token context window still holds roughly 200,000 words: a couple of full books, or a very large code module with room to spare.
Like its siblings it supports tools and reads pictures and video, and it is covered by zero-data-retention terms. That combination — careful reasoning, multimodal input, a manageable footprint and Apache-2.0 weights — makes it a strong pick for people who find the fast tier a bit glib and the trillion-parameter tier a bit excessive. It is the model you choose when the answer matters more than the seconds.
What can you do with Qwen3.8 27B?
Where a careful mid-size model beats a fast one:
- Code review and design questions where being right matters more than being quick.
- Explaining subtle behaviour in code or prose, with the full 262k-token context — roughly 200,000 words — in view.
- Reading screenshots, diagrams and video frames and reasoning carefully about what is in them.
- Multi-step tool workflows that need a steadier hand at each hop.
- Writing tasks where a first draft has to be close to a final draft.
- Homework-style reasoning: maths, logic and analysis problems worked through step by step.
How does Qwen3.8 27B compare?
Next to Qwen3.8 Flash, the 27B is the specialist: Flash is about volume and speed at $0.47 per million tokens with a million-token window, while the 27B is about care at $3.20 with a 262k window. Next to the Qwen3.8 2.4T, it is the sensible middle — a fraction of the price of Alibaba's heavyweight, with most of the family's reasoning character intact. Plenty of people run a Flash model as their act model and the 27B as their think model.
Across makers, its nearest neighbours are other careful mid-size models: Google's Gemma 4 31B writes beautifully and reads images at $0.40, and Meta's Muse Glimmer 30B is an agentic 30B at $1.50. If your work is mostly reasoning over text and images rather than driving tools, the 27B's combination of care and context is hard to beat at this size.
Is Qwen3.8 27B right for you?
Choose the 27B when answer quality is the point: code review you intend to merge on, explanations you intend to rely on, analysis where a hallucination costs more than the tokens. Its dense design and visible thinking make it behave like a careful colleague rather than a fast assistant, and the 262k window comfortably holds most real documents. It is also the Qwen for people who self-host — 27B dense under Apache 2.0 is about the most ownership-friendly serious model on the site.
Skip it when volume is the point: Flash does the same family's speed tier at a sixth of the price, and the difference shows up in invoices long before it shows up in answers. Skip it, too, when a task needs true heavyweight reasoning — the 2.4T sibling is the family's ceiling, at a price to match. The 27B's specialty is the middle of that range, done carefully, with pictures and video read natively along the way.
Frequently asked questions
What is the context window of Qwen3.8 27B?
262,144 tokens — roughly 200,000 words. Smaller than the million-token members of its family, but enough for a couple of full books, a large code module, or very long conversations without any squeezing. Between the Flash tier’s million tokens and this, the difference is mostly about how much history you need — most real documents fit comfortably in either.
Is Qwen3.8 27B a reasoning model?
Yes. It is described as thinking before it answers, and on UsingOpen it ships with a low default reasoning effort — enough deliberation to be careful, not so much that answers crawl. You can ask for more careful treatment on hard problems. The low default effort is well judged for a mid-size model: enough thinking to catch its own mistakes, not so much that a simple question pays a reasoning tax.
Is Qwen3.8 27B good for coding?
Yes — coding and careful answers are its stated remit. It is the Qwen3.8 you pick for review, design discussion and explanation, as opposed to the Flash tier's high-volume generation. The pair to run it against is Qwen3.8 Flash — draft with the fast one, review with this one, and the family’s shared style keeps the hand-off smooth.
Does Qwen3.8 27B support image input?
Yes. It reads pictures and video as well as text, so visual material can join the conversation alongside code and prose.
Who makes Qwen3.8 27B?
Alibaba, through its Qwen team. The Qwen family is one of the most widely adopted open-weight lines in the world and is released under Apache 2.0, which makes the 27B straightforward to self-host or build on.
How does Qwen3.8 27B compare to Qwen3.8 Flash?
The 27B is dense and careful: every parameter is active on every token, it thinks before answering, and it costs $3.20 per million tokens with a 262k window. Flash is sparse and quick at $0.47 with a million-token window. Care versus speed is the whole story. The price multiple is the honest summary: Flash costs a sixth as much, and the 27B earns it only when care is the product.
Is Qwen3.8 27B free to chat with on UsingOpen?
Yes. Pick it in the model picker and start chatting — no account, no payment. History stays on your device, and the picker shows live availability.
Related models on UsingOpen
Qwen3.8 27B sits in busy company — the closest alternatives on UsingOpen are:
- Qwen3.8 Flash by Alibaba: the quick multimodal workhorse of the Qwen3.8 family, with a million-token window built for agent work.
- Gemma 4 31B by Google: the dense 31B Gemma — strong writing, picture input and visible thinking.
- Muse Glimmer 30B by Meta: the compact agent from Meta Superintelligence Labs, distilled to run on ordinary hardware.
- Qwen3.8 2.4T by Alibaba: the open heavyweight — 2.4T parameters with 95B awake for the hardest reasoning.
Head-to-head comparisons
See how Qwen3.8 27B lines up with the models people most often weigh it against:
- Qwen3.8 27B vs Gemma 4 31B — A careful Qwen against Google's budget writer and image reader.
Model availability changes as upstream capacity comes and goes — the picker on the chat page shows what is ready right now.
Chat with Qwen3.8 27B free on UsingOpen — no account, no sign-up
← All open-weight models · Compare models · Free AI chat · FAQ