Qwen3.5-27B vs Qwen3.8 27B
Qwen3.8 27B and Qwen3.5-27B score 69.5% vs 64.1% averaged across four Vision Evals tasks (#28 and #36 of 69 models), so Qwen3.8 27B leads overall.
- Qwen3.8 27B wins 3 of the 4 tasks, with its widest lead on object detection (65.7% vs 50.5%).
- Qwen3.8 27B is about a fifth of the cost: $0.0009 vs $0.0043 overall average cost per sample.
- Qwen3.8 27B is faster: 18.0s vs 80.4s per sample.
Roboflow Vision Evals. Full breakdown below.
Compare Qwen3.5-27B vs Qwen3.8 27B live
Run the same image across every model that supports a task and compare their outputs side-by-side.
Detect and compare bounding boxes across models on the same image.
Upload an image
Drag and drop an image here, or click to browse
Qwen3.5-27B vs Qwen3.8 27B on Vision Evals
Roboflow Vision Evals. Full breakdown below.
Qwen3.5-27B vs Qwen3.8 27B Comparison Table
Evals updated November 12, 2026Pricing updated November 13, 2026
| Property | Qwen3.5-27B | Qwen3.8 27B |
|---|---|---|
| Organization | Qwen | Qwen |
| Category | open | open |
| Modality | multimodal | multimodal |
| Release Date | Feb 2026 | Aug 2026 |
| Context Window | 262K | 262K |
| Parameters | 27B | 27.78B |
| License | Apache 2.0 | Apache 2.0 |
| Commercial use | free for commercial use | free for commercial use |
| Pricing per 1M tokens | ||
| Input $/1M | $0.260 | $0.150 |
| Output $/1M | $2.60 | $1.88 |
| Vision Tasks | ||
| Model Features | ||
Both support 3 more features | ||
Vision Evalsground-truth scores across 5 vision tasks, pooled at low effort | ||
| Overall | 64.1%* | 69.5%* |
| Quantizationsself-hosted | ||
| Avg cost / sample | $0.0043 | $0.0009 |
| Avg speed / sample | 80.37s | 17.99s |
| By task | ||
| Object Detection (low) | 50.5% ±3.5, Mean of 3 runs, range 46.1 to 53.0 | 65.7% ±1.0, Mean of 3 runs, range 64.6 to 66.5 |
| Object Detection (high) | – | 66.1% ±1.4, Mean of 3 runs, range 64.9 to 67.8 |
| Counting (low) | 67.6% ±1.4, Mean of 3 runs, range 66.2 to 68.9 | 64.9% ±4.1, Mean of 3 runs, range 60.8 to 68.9 |
| Counting (high) | – | 68.0% ±2.0, Mean of 3 runs, range 66.2 to 70.3 |
| Identification (low) | 80.2% ±4.7, Mean of 3 runs, range 75.0 to 84.4 | 85.4% ±4.7, Mean of 3 runs, range 81.3 to 90.6 |
| Identification (high) | – | 87.5% ±3.1, Mean of 3 runs, range 84.4 to 90.6 |
| OCR | – | – |
| by category | ||
| Reasoning (low) | 58.1% ±2.3, Mean of 3 runs, range 55.6 to 60.3 | 62.0% ±2.0, Mean of 3 runs, range 60.3 to 64.2 |
| Reasoning (high) | – | 66.0% ±0.7, Mean of 3 runs, range 65.6 to 66.9 |
Qwen3.5-27B vs Qwen3.8 27B: Overview
Qwen3.5-27B is a multimodal dense hybrid model developed by Alibaba Cloud’s Qwen team and released in February 2026 as a high-precision entry in the Qwen3.5 "Medium" series. Unlike its Mixture-of-Experts (MoE) siblings, the 27B model utilizes a dense architecture combining Gated Delta Networks with a feed-forward structure, activating its full parameter suite for every inference to maximize reliability. The model features a native 262K-token context window, extensible to 1M tokens via YaRN (RoPE scaling), and is released under the Apache-2.0 license.
Optimized for agentic workflows, Qwen3.5-27B employs an early-fusion architecture that treats visual and textual data as a unified stream for deep cross-modal reasoning. While the larger MoE variants in the family lead in raw knowledge benchmarks, the 27B model offers a stable and high-density alternative for structured data extraction and spatial perception, contributing to the Qwen3.5 family’s generational leap in OCR accuracy over the previous Qwen3-VL series.
Qwen3.8-27B is a dense vision-language model of roughly 27.78 billion parameters from Alibaba's Qwen team, built on the architectural foundation established by the Qwen3.5 generation. It takes interleaved text, image, and video input through an integrated vision encoder and produces text, covering document and chart understanding, visual question answering, front-end code generation from visual references, and agentic computer-use workflows. The decoder is a hybrid stack of 64 layers that repeats a pattern of three Gated DeltaNet linear-attention blocks, each followed by a feed-forward block, then one grouped-query full-attention block, so only a quarter of the layers maintain a conventional key-value cache. Native context length is 262,144 tokens.
Post-training targets long-horizon agentic work, and the model exposes per-request thinking control that lets callers switch reasoning traces on or off and choose a reasoning effort level. Multi-token prediction weights are trained alongside the main network for speculative decoding.