Cancel your subscriptions, Ox-Alpha is here! (GLM 5.3 Flash)

What you'll learn

• GLM 5.3 Flash is a mixture of experts model with 320 billion parameters, 18 billion active, at a tenth of its predecessor's price.
• It scores 63.4 on DeepSWE, passing Claude Opus 4.8, which sits at 58.0.
• Cost per completed task is 9 cents, against 95 cents for GPT 5.6 Soul and 3.14 dollar for Claude Fable 5.
• During the preview Z.ai served 100 billion tokens per day on Chinese AI chips alone, without Nvidia hardware.
• GLM 5.3 Flash averages 47,000 tokens per task, more than double the 20,000 tokens of GPT 5.6 Luna Max.

Topics

Read next

• Writer voert nieuw AI-model in met lagere tokenkosten

  Writer presenteert een kostenbesparing-gericht AI-model gebaseerd op Z.ai's open source GLM-5.2 voor schaalbare bedrijfsimplementatie.

• Zhipu AI lanceert ZCode als goedkoper alternatief voor Claude Code

  Zhipu AI introduceert ZCode met GLM-5.2-model, gericht op complexe codetaken tegen lagere kosten dan concurrenten.

Sources

What is known about this topic outside the broadcast, and where it says so.

• ai.engineer
• creatordb.app
• jukebox.ag
• prograsec.com
• codersera.com
• medium.com
• tosea.ai
• z.ai
• mindstudio.ai
• datacamp.com

Description from the channel

Check out Higgsfield API: https://higgsfield.ai/s/TvYKkQ Join My Newsletter for Regular AI Updates 👇🏼 https://forwardfuture.com My Links 🔗 👉🏻 X: https://x.com/matthewberman 👉🏻 Forward Future X: https://x.com/forwardfuture 👉🏻 Instagram: https://www.instagram.com/matthewberman_ai 👉🏻 Discord: https://discord.gg/u7wTTGWhuJ 👉🏻 Spotify: https://open.spotify.com/show/6dBxDwxtHl1hpqHhfoXmy8 Media/Sponsorship Inquiries ✅ https://bit.ly/44TC45V Links: https://z.ai/blog/glm-5.3-flash https://x.com/OpenRouter/status/2090544970923184269 https://artificialanalysis.ai/ https://x.com/SemiAnalysis_/status/2092623833630998556