ITG NowHome › AI models › Qwen3.8-27B
Qwen3.8-27B
Alibaba (Qwen) · Released 14 Aug 2026 · Open weights
A capable open model small enough to run on a good home computer.
Key facts
- Maker
- Alibaba (Qwen)
- Released
- 14 Aug 2026
- Open weights
- Yes, you can download it and run it yourself
- License
- Apache 2.0: free for almost any use
- Memory (context window)
- about 262 thousand tokens (how much it can read at once)262,144 tokens exactly
- Size
- 27.8B parametersB = billion, T = trillion. Bigger usually needs a stronger computer.
- Official page
- Official page for Qwen3.8-27B (huggingface.co)
Scores
"Independent" = run by an outside group. "Maker's own number" = the company's own claim.
All-roundArtificial Analysis Intelligence Index
34 index points #13 of 17 models scored independent
One combined score from many independent tests of how smart a model is.
setting: xhigh · as of 17 Oct 2026 · source
Chat & writingLMArena Text
1438 Arena score #14 of 15 models scored independent
When real people compare two AI answers side by side without knowing which AI wrote them, which one do they pick more often?
as of 15 Oct 2026 · source
Build websites & appsLMArena WebDev
1593 Arena score #10 of 18 models scored independent
When people ask two AIs to build the same website or app and compare the results, which one do they pick?
as of 17 Oct 2026 · source
CodingLiveBench 2.0 Coding
86.8 score out of 100 #5 of 15 models scored independent
Fresh coding tasks with tests that check the answer, changed often so models cannot memorize them.
setting: highest supported effort · as of 14 Oct 2026 · source
ReasoningLiveBench 2.0 Reasoning
80.0 score out of 100 #12 of 15 models scored independent
Fresh logic and puzzle questions with one right answer, changed often so models cannot memorize them.
setting: highest supported effort · as of 14 Oct 2026 · source
Science questionsGPQA Diamond
89.2% #4 of 6 models scored maker's own number
Very hard science questions in biology, chemistry and physics, written by PhD experts.
as of 17 Oct 2026 · source
Hardest examHumanity's Last Exam (no tools)
30.8% #5 of 6 models scored maker's own number
2,500 expert questions written to be too hard for AI, answered without web search or code.
as of 17 Oct 2026 · source
New puzzlesARC-AGI-2
42.4% #10 of 11 models scored independent
New picture puzzles a model has never seen, so it has to reason instead of remember. Harder for AI than for people.
setting: XHigh, $0.447 per task · as of 18 Oct 2026 · source
SpeedOutput speed
41 tokens per second #15 of 16 models scored independent
How many tokens (pieces of words) per second the model writes once it starts answering. Higher is faster.
setting: xhigh · as of 17 Oct 2026 · source
How it compares
The 3 models closest to Qwen3.8-27B on the Artificial Analysis Intelligence Index, an all-round score. Qwen3.8-27B: 34 index points, #13 of 17 models scored.
News about Qwen3.8-27B on ITG Now
No stories about Qwen3.8-27B on ITG Now yet.
FAQ
Is Qwen3.8-27B free and open?
Qwen3.8-27B has open weights: you can download it and run it on your own computer or server. License: Apache 2.0: free for almost any use.
How much does Qwen3.8-27B cost?
There is no listed price per token. It has open weights, so you can run it yourself or through a host you pick.
What is Qwen3.8-27B best at?
A capable open model small enough to run on a good home computer. Its best result: #4 of 6 models scored on GPQA Diamond (science questions).
Who makes Qwen3.8-27B?
Qwen3.8-27B is made by Alibaba (Qwen). It came out on 14 Aug 2026.