GLM 5.2 Absent from Fast Inference Platforms as Groq and Cerebras Lag Behind
Neither Groq nor Cerebras has added GLM 5.2; Cerebras is on GLM-4.7, Groq still runs Llama 3.x and gpt-oss.
Developer Simon Willison flagged on X that GLM 5.2 has yet to land on the fastest custom silicon inference platforms. As of June 19, Cerebras is running GLM-4.7 while Groq's catalog remains centered on Llama 3.x and gpt-oss models — neither provider has integrated the newer GLM release.
The gap matters for teams that optimize for throughput and latency. Groq and Cerebras have built their value proposition around inference speeds that can be an order of magnitude faster than standard GPU cloud offerings, making them attractive for real-time and latency-sensitive workloads. When a recently released model isn't yet available on those platforms, developers face a direct tradeoff: run the latest version on conventional infrastructure at higher latency, or stay on an older model version to preserve response-time guarantees. Willison's observation reflects a recurring friction as the pace of model releases continues to outrun the time it takes fast-inference providers to validate and deploy new weights.