Z.ai's GLM-5.3 with max reasoning scores 60 on the Artificial Analysis Intelligence Index, on par with Kimi K3 but below Opus 5 at 63 and Fable 5 at 62
@artificialanlys
Context & Ripple Effects
Z.ai’s GLM line has moved quickly up the same benchmark: GLM-5.2 led the open-weights field at 51 in June, after GLM-5 was introduced as the company’s flagship open-weight model. The new result places GLM-5.3 in the top cluster rather than merely ahead of other open-weight entries.
Z.ai gains a benchmark result that materially improves on GLM-5.2’s 51, strengthening the GLM family’s positioning in reasoning-focused comparisons.
Second-order effects
Opus 5 and Fable 5 retain the benchmark lead, but the narrower spread makes subsequent index updates more consequential for model ranking and comparative claims.
Kimi K3 now shares a directly comparable score with GLM-5.3, putting the two models in the same performance tier on this index rather than leaving Z.ai with a distinct open-weights lead.
Third-order effects
If score gains continue at this pace, headline benchmark leadership will become less durable, shifting competition toward repeated releases and independent ranking updates rather than a single model launch.
The trend: Frontier-model competition is tightening as successive releases rapidly compress gaps on third-party intelligence benchmarks.
GLM-5.3 API is now live. - Built for coding, defensive cybersecurity, and long-horizon agentic tasks - Priced the same as GLM-5.2 - Available via the official API and partner model gateways Get started: https://docs.z.ai/...
@artificialanlys
@artificialanlys on x
Full results across the Artificial Analysis Intelligence Index:
@artificialanlys
@artificialanlys on x
GLM-5.3 improves in AA-Omniscience (4 to 14), primarily driven by an increase in accuracy (23% to 34%)
@artificialanlys
@artificialanlys on x
GLM-5.3 achieves an ELO of 1770 from 1524 to 1770, a 246-point jump from its predecessor.
@openrouter
@openrouter on x
GLM-5.3 from @Zai_org is live on OpenRouter! The same base model as GLM-5.2, with gains entirely from post-training: Terminal-Bench 3.0 jumps from 4.6 to 28.3, and DeepSWE v1.1 from 46.2 to 66.9. Use it now: https://openrouter.ai/...
@teortaxestex
@teortaxestex on x
Phenomenal But, as great as it is, they are overfocusing on SWE a little. Slight regression on CritPt. Kimi K3 is still the holistically strongest Chinese model.
@zixuanli_
Zixuan Li on x
GLM-5.3 scores 60 on the Artificial Analysis Intelligence Index. While GLM is best known for its coding capabilities, its strengths extend far beyond coding. GLM-5.3 delivers significant improvements in reasoning, general chat, and specialized domains such as law and finance.
@designarena
@designarena on x
BREAKING: GLM-5.3 by @Zai_org places 3rd overall on Design Arena with an Elo of 1351. This is a 6-position improvement from GLM-5.2, and makes GLM-5.3 the 2nd-highest-ranked open-weight model on real-world design tasks. Congratulations to the team on the launch!
@ed3d.net
@ed3d.net on bluesky
I think GLM-5.3 is fine. I also think it might be overcooked. It talks like Opus 5 and that is a bad thing. I also have had more off-task sessions with it as my top orchestrator than 5.2. — My working assumption is that as a reviewer and a planning oracle it is very substitu…
@ed3d.net
@ed3d.net on bluesky
AA's numbers are generally pretty well matched to the relative vibes, except that they really overrate frontier midrange models (Terra, Sonnet) — They do not include tests for long-haul directional adherence, though, which is a major gap — You will not notice in practice a 5-…