Some people are mixing up Qwen 4 and the plans for Qwen 4.5 - 5 so let’s clear…
This is a AI post classified by Jev as Frontier models (a model release), kept by the AI Radar because it carries real work, not commentary.
Qwen 4 is coming soon!! Some people are mixing up Qwen 4 and the plans for Qwen 4.5 - 5 so let’s clear everything up! Alibaba’s released configuration already contains `qwen4_exp` and `Qwen4ExpForConditionalGeneration`. These identifiers are in its public Qwen4 architecture preview, released August 26. That’s not the fun stuff, we already know it’s going to be a hybrid linear/sparse attention, gated residual streams, N-gram lookup tables that can be offloaded to system RAM, and Muon-based training. The juicy info was from today’s Apsara Conference, the community has already reported • Q
Posted by Chris (66.6k followers) 52 min ago · 35 likes · 1.7k views · view the original post on X. Kept by the AI Radar as Frontier models.
More AI work like this
- GLM-5.3-FlashX is now officially available in Open Design. — @OpenDesignHQ
- Today's boutta be a movie — @vikktorrrre
- GROK 4.7 xHIGH JUST ENTERED THE TOP AI MODEL RACE. — @ZentrixHQ
- 测试了 4 个在近期发布的模型的鹈鹕骑车：Step 5、Grok 4.7、MiMo V2.6 Flash、MiMo V2.6 Pro — @Fei2411
- We tested GLM 5.3 and Grok 4.7 with the same prompts to build immersive 3D temple… — @OpenDesignHQ
- This day starts amazing: Qwen 4 announced, coming soon. Even our beloved 27b. — @kimmonismus
- 🚨 GPT-6 Sol is showing up in the served model logs — @Mr_Salio
- Grok 4.7 tokens cost 5x less for input and 8.33x less for output than GPT-6 Astra. — @DanDr1s
Every post is read and classified by Jev (TypeSafe): what it is, which market it belongs to, and whether the link is a real tool. 41.3k posts from 5k X accounts over the last 14 days, 1.7k tools, 19 markets. Collected every 5 minutes, fully re-ranked every hour — last update 2026-09-22 11:16 UTC. Full method.