Muhammad Afzal

Building IoT + AI solutions | @MakerBuddycc (easy STEM IoT kits) | (MCP Directory) | Sharing practical tools & projects | Let's build together!

[Muhammad Afzal @AfzalBuilds]

Today i have tested Claude Opus 4.8 VS Codex GPT 5.5 VS GLM 5.2 vs MiniMax M3 to revamp our website. I Provided the same prompt but GLM 5.2 out perform in design beside it lacks in vision capabilities.

Complete video is coming soon, so stay tuned.

[Z.ai @Zai_org]

Introducing GLM-5.2: Frontier Intelligence, Open Weights

- Significant improvements in coding and agentic tasks
- Strong long-horizon capabilities with a 1M context window
- Two levels of reasoning effort: GLM-5.2 (max) pushes the limits, while GLM-5.2 (high) strikes a strong balance between performance and token efficiency
- MIT-licensed open weights
- Same API pricing as GLM-5.1

Tech Blog: https://t.co/LAsxUdN0JZ
Weights: https://t.co/g0A1C4UWx4
API: https://t.co/Kc3E22cbN7
Coding Plan: https://t.co/Nk8Y98HNhU
Chat: https://t.co/WCqWT0qCQb

[Muhammad Afzal @AfzalBuilds]

@cyrilXBT Checkout my comparison where GLM 5.2 is outperforming beside laking in vision capabilities. detailed video is coming soon at my profile so stay tuned.

[Muhammad Afzal @AfzalBuilds]

@hqmank Here is my testing results and GLM 5.2 is giving the tough time to Opus 4.8 as well GPT 5.5

[Muhammad Afzal @AfzalBuilds]

@israfill i tried but it's not working properly many timeouts and slow speed.

[Muhammad Afzal @AfzalBuilds]

@zensidotdev hows it's limits if we compare with official https://t.co/sb2AFfAgke plans?

[Muhammad Afzal @AfzalBuilds]

@nutlope it's insane. in my last video i just focused on reasoning as well as for back-end development but after watching your demo i am thinking to record another review video about GLM 5.2 use case in design.

[Muhammad Afzal @AfzalBuilds]

@HarveenChadha Actually it depends on the requirements but now we have the more options that we can rely on open weight models as compared to proprietary models.

[Muhammad Afzal @AfzalBuilds]

@rileybrown Exactly but i think DeepSeek labs will wait to give bigger shock to everyone ;)

[Muhammad Afzal @AfzalBuilds]

Guys, make your SaaS solutions ready for Agentic Browsing Metric.

What's your results?

[Muhammad Afzal @AfzalBuilds]

@noobnooc it's true it's not cheap as with the launch of GLM 5 series, they instantly increase their prices. And previously they claimed 5x of Claude pro in Lite Plan but now they also removed this line. Actually everyone wants $$$

[Muhammad Afzal @AfzalBuilds]

@markproduct Still ai models didn't have human level critical thinking. In many cases we observed AI agents are good but they are stuck so here humans come up with ideas or solutions and AI gets rid from pooling.

[Muhammad Afzal @AfzalBuilds]

@markproduct Actually still AI needs humans to verify the taste of clients 😜.

[Muhammad Afzal @AfzalBuilds]

@hridoyreh Awesome also try clustova and you will love that.

[Muhammad Afzal @AfzalBuilds]

@ClaudeDevs Thanks dude

[Muhammad Afzal @AfzalBuilds]

@elmoyses @thegenioo Sharing the same context file among multiple ai coding agents.

[Muhammad Afzal @AfzalBuilds]

@Shinnzo_xd @thegenioo Yes opencode go sub is good to go or open router for heavy use.

[Muhammad Afzal @AfzalBuilds]

@jun_song Haha good point ☝️

[Muhammad Afzal @AfzalBuilds]

@jenzhuscott They will immediately raise the prices of their plans 😂

[Muhammad Afzal @AfzalBuilds]

If deepseek is entering into vision then soon many things are going to put the waves in the ocean of AI.

[Jen Zhu @jenzhuscott]

Kids, @deepseek_ai just entered multi-modality mode in apps. ⚡️⚡️⚡️

[Muhammad Afzal @AfzalBuilds]

@CommandCodeAI I am revamping the apps UI/UX using minimax M3