Next AI Model
ArchiveLinesTimeline

Qwen

Alibaba · flagship line · open weights

9 releases since 2023Latest: Qwen3.8-2.4T-A95B

The newest Qwen model is Qwen3.8-2.4T-A95B, released on August 12, 2026. The line's usual gap is about 177 days and 73 have passed, so it is mid-cycle. Alibaba has officially announced Qwen 4; on Sep 22, 2026 it said: “At the conference, Alibaba revealed that its next-generation model, Qwen 4, is currently in training”.

Where the line stands

How long it has been since the last release against the line's usual gap and its recent ones — the rhythm, not a date.

Qwen

Alibaba · flagship line · open weights

Latest: Qwen3.8-2.4T-A95B · Aug 12, 2026

AnnouncedQwen 4 · announced Sep 22, 2026, 32 days ago

“At the conference, Alibaba revealed that its next-generation model, Qwen 4, is currently in training.” — Alibaba, Sep 22, 2026 · Alibaba Cloud

Open-weights release of Qwen 3.8-Max, without the cloud model's image input and non-thinking mode; a commercial licence is required above $50M in annual revenue.

What changed in Qwen3.8-2.4T-A95B →
Days since last release
73
days
Usual gap
177
days
Recent gaps
Qwen3.8-2.4T-A95B · Aug 2026
177
Qwen 3.5 · Feb 2026
210
Qwen3-235B-A22B-Instruct-2507 · Jul 2025
83
Qwen 3 · Apr 2025
222

Get new AI models by email

Every Monday, only when something shipped: the AI models that came out, what each one changed, and which lines are due next.

We email you a link to confirm first · unsubscribe anytime · read past issues · about the newsletter · RSS for the Qwen line

Every Qwen release

Newest first — what each version changed versus the one before, in Alibaba's own numbers.

Aug
12
2026
Qwen3.8-2.4T-A95BAlibaba · QwenOpen weights

Open-weights release of Qwen 3.8-Max, without the cloud model's image input and non-thinking mode; a commercial licence is required above $50M in annual revenue.

Compared with Qwen 3.8-Max · its cloud twin

Scores the same on the all-round intelligence index as its cloud twin Qwen 3.8-Max

Artificial Analysis Intelligence Index v4.3.2 — one score averaging 10 hard tests

40

same as Qwen 3.8-Max (August 3 version)

Same price as its cloud twin Qwen 3.8-Max

Price per 1M tokens — what developers pay, input / output

$2 / $6

same as Qwen 3.8-Max · open weights can also be self-hosted

Sources: Hugging Face · Alibaba Cloud · Artificial Analysis
Feb
16
2026
Qwen 3.5Alibaba · QwenOpen weights
Jul
21
2025
Qwen3-235B-A22B-Instruct-2507Alibaba · QwenOpen weights
Apr
29
2025
Qwen 3Alibaba · QwenOpen weights

Qwen3 open models can switch between step-by-step thinking and quick answers, support 119 languages and dialects, and are licensed under Apache 2.0.

Compared with Qwen 2.5 · the model it replaces

Learned from nearly twice as much text as Qwen 2.5

Pre-training data — how many tokens of text the models learned from, in 119 languages and dialects

about 36 trillion tokens

▲ up from 18 trillion for Qwen2.5 · nearly twice as much

Sources: Qwen · Qwen (GitHub) · Hugging Face
Sep
19
2024
Qwen 2.5Alibaba · QwenOpen weights
Jun
6
2024
Qwen 2Alibaba · QwenOpen weights
Feb
5
2024
Qwen 1.5Alibaba · QwenOpen weights
Nov
30
2023
Qwen-72BAlibaba · QwenOpen weights
Aug
3
2023
Qwen-7BAlibaba · QwenOpen weights