← All models

GLM-ASR

Published Publisher Zhipu AIglm-asr

Model positioning

GLM-ASR is Zhipu AI's speech recognition model. It takes audio as input and produces text as output for speech transcription. The invocation example in the official model documentation uses the API model ID glm-asr. (docs.bigmodel.cn)

Capabilities and uses

  • Uses context to improve the fluency and readability of transcriptions; the official documentation states that it has strong resistance to interference in noisy environments.
  • Supports Chinese, English, and eight regional Chinese dialects: Northeastern Mandarin, Jiao-Liao Mandarin, Beijing Mandarin, Ji-Lu Mandarin, Central Plains Mandarin, Jiang-Huai Mandarin, Lan-Yin Mandarin, and Southwestern Mandarin.
  • Can be used for customer service records, meeting or classroom transcription, subtitle generation, voice interaction, voice search, and other scenarios. (docs.bigmodel.cn)

Limitations

  • Noise resistance does not mean transcription results will be error-free; important content should be checked manually.
  • The current official general speech-to-text API page lists glm-asr-2512 as a model and does not list glm-asr as an option; the original model's current API availability must be confirmed with the platform. GLM-ASR-2512 is a separate version officially announced on December 10, 2025; its specifications or pricing should not be applied directly to this model. (docs.bigmodel.cn)

Pricing

  • Official documentation for mainland China: GLM-ASR is priced at RMB 0.06/minute. The documentation provides an overall per-minute price and does not separately list charges for input audio and output text, nor does it specify billing rounding rules. Official model and pricing page, accessed: 2026-11-18. (docs.bigmodel.cn)
  • The release announcement stated that the model was free for a limited time after launch; this promotion should not be taken as a current commitment to free access. (docs.bigmodel.cn)

Release date and official source

Release date: April 23, 2025. Zhipu AI official model and product release record. (docs.bigmodel.cn)