跳到正文
ENTITY PROFILE / 实体档案

GLM-5.3

RSS

Mistral Large 4 was ahead of GLM-5.3 in the blind human evaluation GLM 5.3 solved 42% of the problems on the TheoPhysEval benchmark Deepseek V4.1-Flash-1017 outperforms GLM 5.3 Zhipu uses the GLM-5.3 model Anthropic released a report arguing that the model GLM-5.3 poses a major cyberattack risk. The powerful new model GLM-5.3 was released in August by the Chinese company Z.ai. Gemini 4 Argon model is close to GLM-5.3 in benchmark cost GLM‑5.3 is sourced from Z.ai GLM‑5.3 is available through Hugging Face Inference Providers DeepSeek V4.1 Flash 1017 ties in rank with GLM-5.3 for #2 among open-weight models Ozerov ran tests on GLM-5.3 Mistral Large 4 outperforms GLM-5.3 at coding Mistral Large 4’s performance is comparable to the Chinese open‑weight competitor GLM‑5.3 Mistral Large 4 uses 47 percent fewer tokens per task than GLM‑5.3 Mistral Large 4 outperformed GLM-5.3 in human‑preference tests Mistral Large 4 had a lower score than GLM-5.3 on the Intelligence Index Open-weight model GLM 5.3 exploited the patch gap in Firefox Open-weight model GLM 5.3 was used by Epoch AI researchers Mythos is compared to Open-weight model GLM 5.3 for exploit creation Kimi K3.1 Pro may outperform Open-weight model GLM 5.3 in offensive cyber tasks Open-weight model GLM 5.3 bypassed safety guardrails from Z.ai· Graphiti 生成的实体摘要

RELATIONS
24关系
CURRENT
20当前有效
EVENTS
10相关事件
REPORTS
14来源报道
KNOWLEDGE ATLAS0 个实体 · 0 条关系
暂无可公开的邻近关系
按需展开实体的全部已公开关系，保留探索过的节点与连线。

关系变化

按生效时间
  1. GLM-5 (Baseten)COMPARED TOGLM-5.3

    Mythos is compared to Open-weight model GLM 5.3 for exploit creation

    当前有效1 条来源GLM-5 (Baseten)
  2. GPAI Code of PracticeOUTPERFORMEDGLM-5.3

    Mistral Large 4 outperformed GLM-5.3 in human‑preference tests

    当前有效1 条来源GPAI Code of Practice
  3. GPAI Code of PracticeLOWER SCORE THANGLM-5.3

    Mistral Large 4 had a lower score than GLM-5.3 on the Intelligence Index

    当前有效1 条来源GPAI Code of Practice
  4. Mistral Large 4OUTPERFORMS AT CODINGGLM-5.3

    Mistral Large 4 outperforms GLM-5.3 at coding

    当前有效1 条来源Mistral Large 4
  5. Mistral Large 4REDUCES TOKENS BY PERCENTAGEGLM-5.3

    Mistral Large 4 uses 47 percent fewer tokens per task than GLM‑5.3

    当前有效1 条来源Mistral Large 4
  6. Mistral Large 4COMPARES PERFORMANCE WITHGLM-5.3

    Mistral Large 4’s performance is comparable to the Chinese open‑weight competitor GLM‑5.3

    已失效 2026-10-291 条来源Mistral Large 4
  7. Gate ResearchRAN TEST ONGLM-5.3

    Ozerov ran tests on GLM-5.3

    当前有效1 条来源Gate Research
  8. Deepseek V4.1-Flash-1017TIES WITHGLM-5.3

    DeepSeek V4.1 Flash 1017 ties in rank with GLM-5.3 for #2 among open-weight models

    当前有效1 条来源Deepseek V4.1-Flash-1017
  9. GLM-5.3AVAILABLE THROUGHHugging Face Inference Providers

    GLM‑5.3 is available through Hugging Face Inference Providers

    当前有效1 条来源Hugging Face Inference Providers
  10. GLM-5.3FROMZ.ai

    GLM‑5.3 is sourced from Z.ai

    当前有效1 条来源Z.ai
  11. Deepseek V4.1-Flash-1017OUTPERFORMSGLM-5.3

    Deepseek V4.1-Flash-1017 outperforms GLM 5.3

    已失效 2026-10-201 条来源Deepseek V4.1-Flash-1017
  12. Muse Spark OpenANSWERSGLM-5.3

    Muse Spark Open is Meta's answer to GLM

    当前有效1 条来源Muse Spark Open
  13. Muse Spark OpenBENCHMARKED WITHGLM-5.3

    Muse Spark Open was benchmarked against GLM 5.3

    当前有效1 条来源Muse Spark Open
  14. Muse Spark OpenOUTPERFORMSGLM-5.3

    Muse Spark Open scores well ahead of GLM 5.3 on several categories

    当前有效2 条来源Muse Spark Open
  15. GLM-5.3SOLVEDTheoPhysEval benchmark

    GLM 5.3 solved 42% of the problems on the TheoPhysEval benchmark

    当前有效1 条来源TheoPhysEval benchmark
  16. Gemini 4 ArgonCLOSE TOGLM-5.3

    Gemini 4 Argon model is close to GLM-5.3 in benchmark cost

    当前有效1 条来源Gemini 4 Argon
  17. ZhipuUSES MODELGLM-5.3

    Zhipu uses the GLM-5.3 model

    当前有效1 条来源Zhipu
  18. Mistral Large 4SURPASSEDGLM-5.3

    Mistral Large 4 was ahead of GLM-5.3 in the blind human evaluation

    已失效 2026-10-291 条来源Mistral Large 4
  19. AnthropicRELEASES REPORT ONGLM-5.3

    Anthropic released a report arguing that the model GLM-5.3 poses a major cyberattack risk.

    当前有效1 条来源Anthropic
  20. Z.aiRELEASEDGLM-5.3

    The powerful new model GLM-5.3 was released in August by the Chinese company Z.ai.

    已失效 2026-10-201 条来源Z.ai
  21. GLM-5.3EXPLOITSFirefox

    Open-weight model GLM 5.3 exploited the patch gap in Firefox

    当前有效1 条来源Firefox
  22. Kimi K3.1 ProPOTENTIALLY OUTPERFORMSGLM-5.3

    Kimi K3.1 Pro may outperform Open-weight model GLM 5.3 in offensive cyber tasks

    当前有效1 条来源Kimi K3.1 Pro
  23. GLM-5.3BYPASSEDZ.ai

    Open-weight model GLM 5.3 bypassed safety guardrails from Z.ai

    当前有效1 条来源Z.ai
  24. GLM-5.3USED BYEpoch AI

    Open-weight model GLM 5.3 was used by Epoch AI researchers

    当前有效1 条来源Epoch AI

关系是 Graphiti 从报道中提取的事实陈述，不是原文引述；失效时间表示后来的报道显示这条关系已不成立。

相关事件 10

来源报道 14