Z.ai releases GLM-5.1, a 754B-parameter model that it says outperforms GPT-5.4 and Claude Opus 4.6 on SWE-bench Pro, available under an MIT license (Carl Franzen/VentureBeat)
Why it matters: Z.ai's GLM-5.1, if claims hold, could shift the competitive landscape for leading AI models like GPT-5.4.
- Z.ai (Zhupai AI) launched GLM-5.1, a 754B-parameter model, asserting it outperforms GPT-5.4 and Claude Opus 4.6 on SWE-bench Pro, and is available under an MIT license (Carl Franzen/VentureBeat).
- Anthropic announced its Mythos Preview model achieved 93.9% on SWE-bench Verified and 77.8% on SWE-bench Pro, significantly higher than Opus 4.6's 80.8% and 53.4% respectively (Michael Nuñez/VentureBeat).
- GLM-5.1's release under an MIT license by a Chinese startup raises questions about China's renewed focus on open-source AI development (Carl Franzen/VentureBeat).
Chinese AI startup Z.ai (Zhupai AI) has released GLM-5.1, a 754B-parameter model, claiming it surpasses GPT-5.4 and Claude Opus 4.6 on the SWE-bench Pro benchmark and is available under an MIT license. This move by Z.ai suggests a potential resurgence of open-source AI initiatives from China, while Anthropic separately announced its Mythos Preview model achieved even higher scores on SWE-bench Verified and Pro, outperforming Opus 4.6.
