Qwen3.8 2.4T A95B

Qwen3.8 2.4T A95B is a model released under Alibaba's Qwen line, the company's series of general-purpose language models. Its name follows the sparse mixture-of-experts naming convention used across recent Qwen releases, where the first figure refers to total parameter count (2.4T) and the "A" figure refers to the parameters active per token (95B). That structure is intended to give a model a very large total capacity while keeping the per-token compute cost closer to that of a mid-to-large dense model. Our database currently holds limited verified specifications for this model. The measurements we do track come from Artificial Analysis, which recorded a median output speed of about 46.9 tokens per second and a time to first token of roughly 1,790 milliseconds. Those figures place it in the range typical of large, high-capacity models rather than small latency-optimised ones: responses stream steadily, but the initial delay before the first token arrives is close to two seconds. Throughput and latency both vary by serving provider, hardware, and load, so the numbers above are best treated as one reference point rather than a guarantee. We do not currently track a confirmed context window, modality list, or reasoning-mode configuration for Qwen3.8 2.4T A95B, so buyers evaluating it for a specific workload should confirm those details against the serving provider's own documentation. Compare the providers listed in the pricing table on this page, since serving cost and measured speed for the same model can differ substantially between hosts.