Genuinely impressed, almost shocked, at how good GLM-5.2 by @zai_org is at coding. This changes things.
7:11 PM · Jun 20, 2026 · 192.7K Views
Next.js creator Guillermo Rauch praises GLM-5.2, saying its coding capabilities could shift software development expectations
Story Overview
Vercel CEO and Next.js creator Guillermo Rauch singled out GLM-5.2 for coding performance that surprised him enough to question current assumptions about AI tools, spotlighting a new open-weight model from Beijing-based Z.ai that rolled out first to paid subscribers before hitting Hugging Face under an MIT license.
1642.8K123626228.7K
Developer Impact
Agentic features arrive ready for repo-scale work
The model ships with a usable million-token context, two thinking-effort modes, and explicit positioning for long-horizon coding agents, letting developers plug it into tools like ZCode or third-party platforms without regional locks.
Open Question
Independent scores remain thin on the ground
No official benchmarks accompanied the launch, leaving the strength of its coding claims to user reports and third-party tests that have so far surfaced only scattered results rather than head-to-head numbers.
Sentiment
Many users expressed excitement about GLM-5.2's strong coding performance after the Vercel CEO highlighted it as shockingly good, while a few raised concerns over speed and competition from other models.
Pos
94.6%
Neg
5.4%
42 comments with sentiment.
Cluster Engagement
Digg Deeper
No Digg Deeper questions have been answered for this story yet.
Posts from X
Most Activity
Most Activity
VIEWS4.4KLIKES32
Miles Brundage@Miles_Brundage
Many are saying
Guillermo Rauch@rauchg
Genuinely impressed, almost shocked, at how good GLM-5.2 by @zai_org is at coding. This changes things.
3hViews 4.4KLikes 32Bookmarks 5
BOOKMARKS15
Guillermo Rauch@rauchg
@Michmue2 @Zai_org from http://vercel.com/ai-gateway
4hViews 2.3KLikes 17Bookmarks 15
RETWEETS2
Jun Song@jun_song
He is CEO of @vercel.
Guillermo Rauch@rauchg
Genuinely impressed, almost shocked, at how good GLM-5.2 by @zai_org is at coding. This changes things.
3hViews 34.6KLikes 353Bookmarks 81
REPLIES3
Jun Song@jun_song
@rauchg @Zai_org Open source will win
3hViews 1.2KLikes 5
Guillermo Rauch@rauchg
@uAghazadae @Zai_org should work perfectly with http://eve.dev / http://ai-sdk.dev
4hViews 1.9KLikes 9Bookmarks 5
uaghazade@uAghazadae
@rauchg @Zai_org does the ai sdk include this?
4hViews 2KLikes 3Bookmarks 1
mike@Michmue2
@rauchg @Zai_org what hardware setup are u using?
4hViews 2.2KLikes 4
Concentrate AI@concentrateai
@osaki_io @rauchg @Zai_org https://concentrate.ai/docs/integrations/codex
2hViews 15Bookmarks 1
OpenModels@openmodels_
@rauchg @Zai_org same reaction, GLM-5.2 is surprisingly strong at coding
we offer it on OpenModels with 1M context, $1.182/M input + $4.136/M output, and pay-as-you-go credits
very compelling for longer agent/coding runs
4hViews 457Likes 5
Robert Scoble@Scobleizer
@jun_song @vercel Thank you for pointing that out, because it makes it a thousand times more important to pay attention to.
1hViews 124Likes 2
dhru@dhruzzz
@rauchg @Zai_org when coming on aigateway ?
3hViews 387Likes 1
Emon Datta@emonuxui
@rauchg @Zai_org Every few months we reset what ‘good at coding’ even means .
3hViews 188
Mike Gannotti@MichaelGannotti
@rauchg @Zai_org It is really impressive
3hViews 673Likes 2
Eclipse 🌖@ECLresearch
@rauchg @Zai_org GLM-5.2’s coding throughput numbers are up 37% vs 4.1 on our internal SWE-bench fork. If that holds against production loads, the tier chart just got a new top right quadrant.
4hViews 607Likes 2
@rauchg @Zai_org
4hViews 504Likes 2
Saint Patrick@SaintPatrick333
@pranavrameshh @jun_song @rauchg @Zai_org Why's that funny?
3hViews 12Likes 1
valx@2Q9OuU619nQL4Nx
@rauchg @Zai_org If it starts writing my tax return, I might finally stop pretending I understand spreadsheets.
3hViews 1.1KLikes 1
Alan@Alan_Earn
@jun_song @vercel have you tried GLM 5.2?
3hViews 22
yash@yashetal
@rauchg @Zai_org Open source and being the frontier of it, deepseek needs to catch-up ig and push some benchmarks
3hViews 575Likes 1
Majid Manzarpour@majidmanzarpour
@rauchg @Zai_org Pretty solid
3hViews 439Likes 1
Load more posts