Qwen4-27B
A large language model in Alibaba's Qwen series with 27B parameters, positioned between small and large models; community speculation holds that it has long-context capability.
Timeline (last 90 days)
The leaker said the Qwen4-27B version will be announced at the Apsara conference in the Netherlands, with the timing pointing to "tomorrow".
A Reddit community post links Qwen4-27B to Pull Request #22 in the GitHub repository analytics-zoo/vllm-xpu-kernels.
The community interpretation holds that the PR implies the KV cache size of Qwen4-27B has not decreased.
All Facts (3)
A Reddit community post links Qwen4-27B to Pull Request #22 in the GitHub repository analytics-zoo/vllm-xpu-kernels.
50%UnverifiedThe community interpretation holds that the PR implies the KV cache size of Qwen4-27B has not decreased.
50%UnverifiedThe leaker said the Qwen4-27B version will be announced at the Apsara conference in the Netherlands, with the timing pointing to "tomorrow".
50%