Roboflow

Qwen3.6 35B A3B vs Qwen3.8 27B

Qwen3.8 27B and Qwen3.6 35B A3B score 69.5% vs 64.8% averaged across four Vision Evals tasks (#32 and #38 of 73 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 57.0%).
  • Qwen3.8 27B is cheaper: $0.0009 vs $0.0012 overall average cost per sample.
  • Qwen3.8 27B is faster: 18.0s vs 27.1s per sample.

Compare Qwen3.6 35B A3B 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.

Open Object Detection in the full playground
QwenQwen3.6 35B A3B
Run to compare this model.
QwenQwen3.8 27B
Run to compare this model.

Qwen3.6 35B A3B vs Qwen3.8 27B on Vision Evals

Score by taskQwen3.6 35B A3BQwen3.8 27B
Object Detection
57.0
65.7
Reasoning
54.8
62.0
Identification
82.3
85.4
Counting
65.3
64.9

Roboflow Vision Evals. Full breakdown below.

Qwen3.6 35B A3B vs Qwen3.8 27B Comparison Table

Evals updated November 19, 2026Pricing updated November 22, 2026

PropertyQwen3.6 35B A3BQwen3.8 27B
OrganizationQwenQwen
Categoryopenopen
Modalitymultimodalmultimodal
Release DateApr 2026Aug 2026
Context Window262K262K
Parameters35B total, 3B active27.78B
LicenseApache 2.0Apache 2.0
Commercial usefree for commercial usefree for commercial use
Pricing per 1M tokens
Input $/1M$0.100$0.035
Output $/1M$0.900$1.50
Vision Tasks
Phrase GroundingSupportedNot listed
Video ClassificationSupportedNot listed
Both support 10 more tasks
Model Features
Both support 3 more features
Vision Evalsground-truth scores across 5 vision tasks, pooled at low effort
Overall
64.8%*
69.5%*
Quantizationsself-hosted
FP864.8%AWQ-INT460.9%hardware →
BF1669.5%FP868.7%AWQ-INT469.5%hardware →
Avg cost / sample$0.0012$0.0009
Avg speed / sample27.10s17.99s
By task
Object Detection (low)
57.0%
±1.3, Mean of 3 runs, range 56.1 to 58.7
$0
65.7%
±1.0, Mean of 3 runs, range 64.6 to 66.5
$0
Object Detection (high)–
66.1%
±1.4, Mean of 3 runs, range 64.9 to 67.8
$0
Counting (low)
65.3%
±2.7, Mean of 3 runs, range 62.2 to 67.6
$0
64.9%
±4.1, Mean of 3 runs, range 60.8 to 68.9
$0
Counting (high)–
68.0%
±2.0, Mean of 3 runs, range 66.2 to 70.3
$0
Identification (low)
82.3%
±6.3, Mean of 3 runs, range 75.0 to 87.5
$0
85.4%
±4.7, Mean of 3 runs, range 81.3 to 90.6
$0
Identification (high)–
87.5%
±3.1, Mean of 3 runs, range 84.4 to 90.6
$0
OCR––
by category
Reasoning (low)
54.8%
±1.3, Mean of 3 runs, range 53.0 to 55.6
$0
62.0%
±2.0, Mean of 3 runs, range 60.3 to 64.2
$0
Reasoning (high)–
66.0%
±0.7, Mean of 3 runs, range 65.6 to 66.9
$0

Qwen3.6 35B A3B vs Qwen3.8 27B: Overview

Qwen3.6 35B A3B

Qwen3.6-35B-A3B is a sparse Mixture-of-Experts (MoE) multimodal language model developed by the Qwen team at Alibaba Group. It carries 35 billion total parameters but activates only approximately 3 billion per forward pass via a learned routing mechanism, giving it the representational capacity of a large dense model at a fraction of the inference compute. The model is natively multimodal, processing images, documents, and video alongside text as a core architectural capability rather than an add-on. It supports a native context window of 262,144 tokens, extensible up to 1,010,000 tokens via YaRN. A key design feature is the unified thinking/non-thinking mode framework: users can switch between deliberate chain-of-thought reasoning and fast direct responses within a single model, and a "thinking preservation" option retains reasoning context across multi-turn agentic workflows to reduce redundant computation.

The model is specifically optimized for agentic coding tasks, including repository-level reasoning, frontend workflow generation, multi-step tool use, and MCP (Model Context Protocol) integration. For vision-language tasks it achieves 92.0 on RefCOCO, 89.9 on OmniDocBench 1.5, and 83.7 on VideoMMMU. The model also supports Multi-Token Prediction (MTP) for speculative decoding. All Qwen3.6 open-weight models are released under the Apache 2.0 license.

Qwen3.8 27B

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.