glm 5.2 already better than fable, which is 1/5th the cost of kimi 3 currently... so will have to wait for price compression when it goes open to find out
Vibes I'd quantify as what you can actually use. Most Fableposts I see are about fighting with guardrails and slowness, costs or useage limits.
To the extent there are positive vibes around Anthropic generally I think is more of a cult or fad mentality. There's a predisposition around left-aligned people towards it, low information users that simply recognize the name, people that like paying a premium since they equate that with status, or people that buy-into Dario's FUD / gov reaction and interpret that as achievement.
I've been Composer maxxing for months for example, is it the leading general intelligence? No, but it's been the most usable. Been leaning on Grok 4.5 more since its really cheap for Cursor users. I do have access to Anthropic models via Cursor and try them occasionally for sanity checks and blocker busting, and I always regret it as a waste of credits. Just the other day Sonnet 5 imported a dep from a local folder in another project that wasn't even in the workspace, and that slop probably cost me the equivalent of 3 days worth of Grok/Composer credits.
I've haven't yet regretted GLM for sanity checks or tearing down a blocker. I've even started using a different editor to get GLM even cheaper, Anthropic has never justified modifying my workflow.
On the other hand, I know a few devs moving to Cursor or OpenCode from the Claude stack, guys that have been Anthropic maxxing for awhile because it's been the consensus trendy option.... only to realize they're getting scammed. Those are analogous workflow changes not because something else is good, but because Anthropic is bad.
glm 5.2 already better than fable, which is 1/5th the cost of kimi 3 currently... so will have to wait for price compression when it goes open to find out
I'd give you better than Opus 4.7 or better than GPT5.5. But better than fable or GPT5.6 probably no. That's exaggerating it.
Better being subjective of course, so assessment generally boils down to benchmarks vs. vibes.
In synthetic benchmarks they're typically neck and neck, plenty of indie benchmarks where GLM wins though.
This Kilo Code post was interesting: https://x.com/kilocode/status/2067915121532325929?s=20
Design Arena had a good post as well where it won:
https://medium.com/data-science-in-your-pocket/glm-5-2-beats-claude-fable-5-glm-5-2-benchmarks-explained-493751c8a24f
Vibes I'd quantify as what you can actually use. Most Fableposts I see are about fighting with guardrails and slowness, costs or useage limits.
To the extent there are positive vibes around Anthropic generally I think is more of a cult or fad mentality. There's a predisposition around left-aligned people towards it, low information users that simply recognize the name, people that like paying a premium since they equate that with status, or people that buy-into Dario's FUD / gov reaction and interpret that as achievement.
I've been Composer maxxing for months for example, is it the leading general intelligence? No, but it's been the most usable. Been leaning on Grok 4.5 more since its really cheap for Cursor users. I do have access to Anthropic models via Cursor and try them occasionally for sanity checks and blocker busting, and I always regret it as a waste of credits. Just the other day Sonnet 5 imported a dep from a local folder in another project that wasn't even in the workspace, and that slop probably cost me the equivalent of 3 days worth of Grok/Composer credits.
I've haven't yet regretted GLM for sanity checks or tearing down a blocker. I've even started using a different editor to get GLM even cheaper, Anthropic has never justified modifying my workflow.
On the other hand, I know a few devs moving to Cursor or OpenCode from the Claude stack, guys that have been Anthropic maxxing for awhile because it's been the consensus trendy option.... only to realize they're getting scammed. Those are analogous workflow changes not because something else is good, but because Anthropic is bad.
https://twiiit.com/kilocode/status/2067915121532325929
https://twiiit.com/kimi_moonshot/status/2077830229968683203