Qwen3.8 27B (Consensus Protocol)

Codersera's guide confirms Qwen3.8-27B as a dense 27B Apache-2.0 vision-language model released on August 14, 2026, with a 262,144-token native context and a 1-million-token extended context that is currently a Qwen Cloud hosted "coming soon" feature rather than a local capability. It reports an Artificial Analysis Int The guide records strong community uptake, with Unsloth GGUF quantizations surpassing 2.7 million downloads, Ollama logging 236,000 pulls, and a Hacker News release thread hitting 1,423 points within four days. It notes a default reasoning effort of "xhigh" causing overthinking issues, describes an FP8 fine-grained qua

DataNorth's news write-up reports that Alibaba released Qwen3.8-27B on August 14, 2026 as an open-weights dense multimodal model under Apache 2.0, accepting text, images, and video with a 262,144-token native context extendable to 1,000,000 tokens. It cites Alibaba's own evaluation scoring the model at 61.7 percent on The article clarifies that Qwen3.8-27B is a native vision-language model rather than a text model with an adapter, with a configuration file that includes a full vision encoder and a language-model-only flag set to false. It also documents an FP8 variant, Qwen3.8-27B-FP8, using fine-grained FP8 quantization with a bloc

Qwen3.8-27B is an Apache-2.0-licensed vision-language model from Alibaba's Qwen team, released on Hugging Face in August 2026. It swapped most attention layers for Gated DeltaNet linear attention, enabling a 262,144-token native context without quadratic memory blow-up. The model is small enough to run on a single high-end GPU or a beefy Mac. According to Qwen's own benchmark tables, Qwen3.8-27B beats Claude Opus 4.6 Max on computer-use and browser-use tasks, though it trails on raw reasoning benchmarks like HLE and GPQA Diamond. It topped Hacker News within a day of release and became the most-liked model on Hugging Face's trending list that week, reflecting strong community adoption.

Kingy.ai's launch-day analysis describes Qwen3.8-27B as a 27.78-billion-parameter dense multimodal model released by Alibaba's Qwen team on August 14, 2026, accepting text, images, and video under Apache 2.0 with a native 262,144-token context window. The review reports Qwen's own benchmark gains over Qwen3.6-27B, incl The same article flags that all launch benchmarks come from Qwen, several are in-house or modified, and the SWE-bench Pro comparison imports Anthropic's Opus result rather than rerunning it under Qwen's setup. It further notes that the 1-million-token context requires YaRN scaling with caveats for shorter prompts, and

OfficeChai reports that Alibaba released Qwen3.8-27B alongside the open weights of Qwen3.8-2.4T-A95B, both under Apache 2.0, framing the 27B model as a local-first alternative to data-center-scale systems. It highlights a native 262K-token context window extendable to 1 million tokens via YaRN and positions the model s The article's benchmark coverage focuses on Alibaba's comparison against Qwen3.6-27B, Qwen3.7-Plus, Meta's Muse Glimmer-30B, and Anthropic's Opus 4.6 Max, with Qwen3.8-27B posting 73.0 versus Muse Glimmer's 51.7 on Terminal Bench 2.1, 79.5 to 77.0 on IFBench, and 89.2 to 83.5 on GPQA Diamond. It notes that Muse Glimmer

Alibaba's Qwen team released Qwen3.8-27B as a post-trained open-weight model on Hugging Face, following the Qwen3.5 and Qwen3.6 series. The card describes it as the most capable Qwen generation, built on Qwen3.5's architecture foundation. It targets coding, professional work, research, and long-horizon agentic tasks with stronger autonomous planning and environment feedback handling. Qwen3.8-27B is a dense 27-billion-parameter causal language model with a vision encoder, supporting native image and video understanding. It ships with flexible thinking control (per-request toggle, reasoning effort tuning, and preserve-thinking for context retention). A managed Qwen Cloud hosted version is teased with 1M context and built-in tools.

Local AI Zone's technical breakdown specifies Qwen3.8-27B as a 27,781,427,952-parameter dense model with 64 Transformer blocks, a hidden dimension of 5,120, 24 query heads with 4 KV heads, a head dimension of 256, and a 248,320-token multilingual vocabulary. It adopts a 3:1 hybrid attention scheme with 48 Gated DeltaNe The same analysis frames Qwen3.8-27B as targeting a 24GB VRAM minimum for local deployment and claims competitive performance versus models 10–15 times its size, including outperforming Meta's Muse Glimmer 30B across 8 direct comparison benchmarks and surpassing Claude Opus 4.6 on 15 of 19 overlapping tests. It positio

The LLM Gateway catalog page for Qwen3.8 27B lists it as a stable, dense 27-billion-parameter Qwen3.8 model released on August 14, 2026, with a 1,000,000-token context window and streaming, tool-calling, reasoning, and soft JSON output support (no strict JSON schema enforcement). Pricing starts at $0.42 per million inp The same page describes Qwen3.8 27B as suitable for coding, professional work, research, and long-horizon agentic tasks, accessed via LLM Gateway's OpenAI-compatible API with automatic provider routing, fallback, and cost analytics. It links to related Alibaba Cloud models including Qwen3.8 Flash, Qwen Image 3.0, Qwen3