Running gpt and glm-5.1 side by side. Honestly can’t tell the difference
So I have been running gpt and glm-5.1 side by side lately and tbh the gap is way smaller than what im paying for
On SWE-Bench Pro glm-5.1 actually took the top spot globally, beat gpt-5.4 and opus 4.6. overall coding score is like 55 vs gpt5.4 at 58. didnt expect that from an open source model ngl
Switching between them during the day I honestly can't tell which one did what half the time. debugging, refactoring, multi-file stuff, both just handle it
GPT still has that edge when things get really complex tho, like deep system design stuff where you need the model to actually think hard. thats where i notice the diffrence
For the regular grind tho it's hard to care about a 3 point gap when my tokens last way longer lol. and they got here stupid fast compared to the 'Thinking' delays which is the part that gets me