Qwen3.8 27B
An open-source lightweight multimodal large language model released by Alibaba's Qwen team, with 27 billion parameters. It supports native multimodal capabilities including text and image understanding, a native context window of 260,000 tokens, and is released under the Apache 2.0 license.
Core Facts
The reviewer believes that Qwen3.8 27B is stronger than the previous generation Qwen3.7 in almost all aspects of coding capability.
The 4-bit quantized weights of Qwen3 27B are only approximately 16.1 GB, allowing it to run on a MacBook Pro/Mac Studio with 32 GB of memory or a consumer-grade GPU with approximately 17 GB of VRAM.
Timeline (last 90 days)
In the second prompt (dialogue about roommates dividing a rice cooker), Hemingway ranked first, HumanLike Chat ranked second, the official original ranked third, and Dongfang Shuo ranked toward the bottom.
In the first prompt, HumanLike Chat ranked second, the official original Qwen3.8 27B ranked third, and DeepSeek v4.1 ranked fifth.
Combining the three prompts, the final ranking was: Hemingway first, HumanLike Chat second, official original Qwen3.8 27B third, DavidAU Brandings fourth, Alexs WinX fifth, DeepSeek v4.1 sixth, and Dongfang Shuo seventh.
The observation noted that models with strong base capabilities (such as the official original) are not necessarily more impressive at literary creation than specially fine-tuned versions, while models focused on anthropomorphic chat instead have more of an advantage in emotional expression.
A Bilibili UP host's comparative review tested seven models with three literary creation prompts: six community fine-tuned versions based on Qwen3.8 27B (Hemingway, HumanLike Chat, Alexs WinX, DavidAU Brandings, Dongfang Shuo) and the official original Qwen3.8 27B, plus DeepSeek v4.1.
The Qwen3.8 27B model scores 34, higher than Beam (501B) at 30 and Nemotron 3.5 Bolt (74B) at 21.
RunInfra does not train models itself; instead it integrates open-source models such as GLM 5.3, DeepSeek V4.1 Flash and Qwen3.8 27B.
The three-tier storage approach is only applicable to MoE models and does not work with Dense models (such as Qwen3.8 27B).
The X-high thinking tier of Qwen3.8 27B consumes an extremely large number of tokens (e.g., 11,000 for drawing a pelican and 35,000 for drawing a dragon), while the medium tier is the best cost-performance sweet spot.
Qwen3.8 27B surpasses Opus 4.6 Max in visual performance on computer use and browser operation tasks.
14 more timeline events
All Facts (20)
The reviewer believes that Qwen3.8 27B is stronger than the previous generation Qwen3.7 in almost all aspects of coding capability.
65%VerifiedThe 4-bit quantized weights of Qwen3 27B are only approximately 16.1 GB, allowing it to run on a MacBook Pro/Mac Studio with 32 GB of memory or a consumer-grade GPU with approximately 17 GB of VRAM.
65%UnverifiedThe observation noted that models with strong base capabilities (such as the official original) are not necessarily more impressive at literary creation than specially fine-tuned versions, while models focused on anthropomorphic chat instead have more of an advantage in emotional expression.
70%UnverifiedIn the second prompt (dialogue about roommates dividing a rice cooker), Hemingway ranked first, HumanLike Chat ranked second, the official original ranked third, and Dongfang Shuo ranked toward the bottom.
50%UnverifiedCombining the three prompts, the final ranking was: Hemingway first, HumanLike Chat second, official original Qwen3.8 27B third, DavidAU Brandings fourth, Alexs WinX fifth, DeepSeek v4.1 sixth, and Dongfang Shuo seventh.
50%UnverifiedIn the first prompt, HumanLike Chat ranked second, the official original Qwen3.8 27B ranked third, and DeepSeek v4.1 ranked fifth.
50%UnverifiedThe Qwen3.8 27B model scores 34, higher than Beam (501B) at 30 and Nemotron 3.5 Bolt (74B) at 21.
50%UnverifiedRunInfra does not train models itself; instead it integrates open-source models such as GLM 5.3, DeepSeek V4.1 Flash and Qwen3.8 27B.
50%UnverifiedThe three-tier storage approach is only applicable to MoE models and does not work with Dense models (such as Qwen3.8 27B).
50%UnverifiedThe ablation version of Qwen3.8 27B is prone to falling into an infinite loop of repetitive decoding during the thinking phase, and the blogger does not recommend using it.
50%UnverifiedThe X-high thinking tier of Qwen3.8 27B consumes an extremely large number of tokens (e.g., 11,000 for drawing a pelican and 35,000 for drawing a dragon), while the medium tier is the best cost-performance sweet spot.
50%UnverifiedThe choice of inference tier has a greater impact on model usability than the choice of quantization method.
50%UnverifiedOn an RTX Pro 6000 (96 GB VRAM), the official FP8 version combined with built-in MTP speculative decoding can reach approximately 120 tokens/second, with quality nearly indistinguishable from 16-bit.
50%UnverifiedSGLang using official NVFP4 weights with D-Spark speculative decoding achieved an average of approximately 173 tokens/second on Blackwell GPUs, approaching 200 for short tasks, while loading a 262k context window.
50%UnverifiedQwen3.8 27B outperforms GLM 5.2 and some GPT 5.6 models on the agentic index.
50%UnverifiedQwen3.8 27B surpasses Opus 4.6 Max in visual performance on computer use and browser operation tasks.
50%UnverifiedThe official Anifer project initially only supported a single RTX 5090, and only five official models were registered, focusing on Qwen3.8 27B, Qwen3.6 27B, and Qwen3.6 35B MoE models.
50%UnverifiedThere is a smaller approximately 9GB version of the Qwen3.8 27B model on Hugging Face, but its quality will be degraded.
50%UnverifiedQwen3.8 27B still makes notable errors in complex 3D modeling, such as scoring only 2 out of 10 on an F1 car drifting scene and producing common-sense violations in SVG artwork.
50%UnverifiedIn a frontend design test, Qwen3.8 27B's context window was limited to 32K under the current configuration, resulting in incomplete code generation with bugs such as abnormal page reflections and missing SVG icons.
50%