Recently tested GLM 4.5 on Claude Code, and the results were surprisingly good.

Used my standard test method: had the AI build a Next.js CRM application requiring SQLite database with API routes connecting frontend and backend.

Better than expected outcomes:

1. Code generation quality approaches Sonnet 4 level

2. Blazing fast – automatically used create-next-app to initialize v15.5.2

3. Fully implemented Prisma ORM with SQLite, even included data seeding

4. Required some prompt tuning but never got stuck

5. Total cost $0.50, even cheaper with @Zai_org subscription

It might be premature to call GLM 4.5 the best open-source programming model, but it's definitely top-tier now.

(Video shows sped-up actual implementation)

PS: DM for referral link – didn't include directly to avoid being too promotional.

---

Reply to comparison queries with Kimi: Haven't tested latest version, but benchmarks show it outperforms Kimi K2. Previously switched back from Kimi to Sonnet.

[Image: Landsat Technologies security solutions ad – unrelated to topic]