Qwen3.8 27B

Qwen3.8 27B is a chat-oriented large language model attributed to Alibaba, the group behind the Qwen model series. The "27B" in the name indicates a parameter count of roughly 27 billion, placing it in the mid-size class of open-weight-style models — larger than the small 4B-to-8B tiers that target cheap, high-volume work, but well below the very large mixture-of-experts models that sit at the top of most vendor lineups. Models in this size band are commonly chosen when a single accelerator or a small multi-GPU node needs to serve general assistant workloads. Our catalog entry for Qwen3.8 27B currently carries limited verified technical detail. We track it as a chat model, and we do not have confirmed figures for its context window, supported input modalities, or tool-calling interface, so this page does not assert values for those fields. Throughput and time-to-first-token measurements sourced from Artificial Analysis are not yet populated for this entry either; where a provider publishes its own latency numbers, those will generally be the more reliable reference for a specific deployment. Readers evaluating the model for production should confirm context length, quantization, and API feature support directly with whichever host they plan to use, since these vary by serving stack even for an identical checkpoint. In practice, a 27B-class Qwen chat model is typically deployed for general conversational assistants, summarization, drafting, and light code assistance, where the tradeoff between response quality and serving cost matters more than absolute peak capability. Compared with much smaller Qwen variants it should handle multi-step instructions and longer prompts more reliably; compared with the largest models in the ecosystem it will generally be cheaper to serve and faster per token, at some cost in reasoning depth. The pricing table on this page shows which providers currently host it and what they charge.