Hangzhou, China · Founded 2023 (Qwen team, Alibaba Cloud)
Alibaba Qwen
The Qwen family — frontier scale at Chinese-cloud prices.
Alibaba's Qwen team ships some of the strongest models outside the US labs. Qwen 3.8 Max beat GPT-5.6 Sol on SWE-bench Pro (67.7 vs 64.6) at launch, and Qwen 3.8 Flash delivers SWE-bench Pro 62.5 from just 6B active parameters at $0.16/1M.
Rankings refresh dailyScored on 6 criteriaNo paid rankings
Qwen 3.8 Max beat GPT-5.6 Sol on SWE-bench Pro at launch (67.7 vs 64.6)
Qwen 3.8 Flash: SWE-bench Pro 62.5 at $0.16/1M — from 6B active parameters
Qwen3.8-Flash-Next open weights preview the Qwen4 architecture
3 models
All Alibaba Qwen Models
Every Alibaba Qwen model in the directory, ranked by overall capability score.
AlibabaBalanced
Qwen 3.8 Max
Alibaba's largest model ever — a 2.4-trillion-parameter MoE (95B active) multimodal flagship that beat GPT-5.6 Sol on SWE-bench Pro and ranks #2 globally for vision.
Verdict
Best Chinese flagship — beats GPT-5.6 Sol on coding, #2 globally for vision.
Quality score
90%
Pricing
$2.00/1M in
$6.00/1M out
Speed
Balanced
3/5 speed
Context
1M tokens
Announced August 3, 2026 on Alibaba Cloud Model Studio; open weights promised a week after launch. $2/$6 is first-party Model Studio pricing; cache reads from $0.17/1M. Announcement moved Alibaba stock +7% in Hong Kong.
Alibaba's first closed-weight flagship — an agent-first model with native extended thinking, built to run autonomously for up to ~35 hours firing thousands of tool calls.
Verdict
Agent-first Qwen flagship, superseded by Qwen 3.8 Max.
Quality score
87%
Pricing
$2.50/1M in
$7.50/1M out
Speed
Balanced
3/5 speed
Context
1M tokens
Announced at Alibaba Cloud Summit May 20, 2026. API-only on DashScope/Model Studio. Cached input $0.25/1M.
Alibaba's preview of the Qwen4 architecture — 125B parameters with only 6B active per token, at sixteen cents per million input.
Verdict
SWE-bench Pro 62.5 at sixteen cents per million input.
Quality score
77%
Pricing
$0.16/1M in
$0.47/1M out
Speed
Very fast
5/5 speed
Context
991k tokens
Released August 26, 2026. The open-weight release is Qwen3.8-Flash-Next, a preview of the Qwen4 architecture: 125B mixture-of-experts with 6B active per token, a 51B n-gram embedding table and a 4B multi-token prediction layer. Qwen 3.8 Flash is the production API version on Qwen Cloud at $0.16/$0.47.
Get notified when Alibaba Qwen releases new models
Pricing changes, new releases, and ranking shifts — straight to your inbox.
No spam. Useful updates only. Affiliate disclosures always clearly labeled.
Alibaba Qwen FAQ
What is Alibaba's best AI model in 2026?
Qwen 3.8 Max ($2/$6) is Alibaba's flagship — a 2.4T-parameter MoE that beat GPT-5.6 Sol on SWE-bench Pro at launch and ranks #2 globally on vision. Qwen 3.8 Flash ($0.16/$0.47) is the value pick: SWE-bench Pro 62.5 from a 125B MoE with only 6B active parameters per token.
Are Qwen models open source?
Partially. Qwen 3.8 Max is API-only — a break from Qwen tradition. But the Qwen3.8-Flash-Next release (August 26, 2026) is open-weight and previews the Qwen4 architecture: 125B mixture-of-experts, 6B active per token, plus a 51B n-gram embedding table.
How does Qwen compare to Claude and GPT?
Qwen 3.8 Max genuinely beat GPT-5.6 Sol on SWE-bench Pro at launch (67.7 vs 64.6), though the Claude frontier — Fable 5 at 80.3% — remains well ahead. Where Qwen wins is price: Max costs a third of Sol, and Qwen 3.8 Flash delivers GLM-5.2-class coding at $0.16/1M input.
Where can I access Qwen models?
Via Alibaba Cloud's Model Studio API (international endpoint available), chat.qwen.ai for consumer use, and — for the open-weight Flash-Next variant — self-hosting or third-party hosts. Data-residency-sensitive teams should note the first-party API routes through Alibaba Cloud.