The Qwen3.5-35B-A3B is a native vision-language model developed by Alibaba Cloud’s Qwen team, released on February 24, 2026, as a high-efficiency entry in the Qwen 3.5 family. It utilizes a sophisticated hybrid architecture that integrates Gated Delta Networks with a sparse Mixture-of-Experts (MoE) system. While the model houses 35 billion total parameters, its routing mechanism activates only 8 routed experts and 1 shared expert per token, totaling approximately 3 billion active parameters. This design achieves cross-generational parity with the previous flagship Qwen3-235B dense model, delivering comparable reasoning and multimodal intelligence with significantly reduced inference latency and compute requirements. Available under the Apache 2.0 license, it is released in both base and instruction-tuned variants for seamless integration with open-source stacks like vLLM and Hugging Face Transformers.
Designed for the emerging era of agentic AI, the model utilizes a unified multimodal foundation built through early-fusion training. This approach allows it to outperform the prior Qwen3-VL series in spatial grounding, document analysis, and UI/GUI interaction. It features a native context window of 262,144 tokens, which is extensible up to 1,010,000 tokensvia RoPE scaling, and provides global support for 201 languages and dialects. This combination of a compact active parameter count and frontier-level visual comprehension makes it a versatile tool for developers requiring a balance of high-throughput speed and sophisticated visual reasoning for long-context workflows.
Drag and drop an image here, or click to browse
Results appear here. Add an image or pick an example to run Qwen3.5 35B A3B.
—
Usage
Past 30 DaysVision Evals is Roboflow's ground-truth benchmark: every model runs the same real-world samples across five vision tasks, and answers are scored against ground truth.
Evals updated October 13, 2026Pricing updated October 13, 2026
Qwen3.5 35B A3B averages 62.5% across the 4 of 5 Vision Evals tasks it has results for, ranking #32 of 63 models overall.
Its weakest relative showing is Identification, ranking #50 of 63 at 80.2%.
At $0.0040 per sample it is the 23rd cheapest of the 63 benchmarked models, and its average inference time of 31.9s per sample makes it the 56th fastest.
Field medians: Object Detection 54.4%, Counting 63.5%, Identification 84.4%, Reasoning 57.6%.
| Task | Score | Field (0 to 100) | Rank | Cost / sample | Speed |
|---|---|---|---|---|---|
| Object Detection | 52.9% ±3.2, Mean of 3 runs, range 49.5 to 55.9 | #34 of 63 | $0.0031 | 24.78s | |
| Counting | 62.6% ±2.0, Mean of 3 runs, range 60.8 to 64.9 | #33 of 63 | $0.0056 | 44.54s | |
| Identification | 80.2% ±1.6, Mean of 3 runs, range 78.1 to 81.3 | #50 of 63 | $0.0032 | 22.22s | |
| Reasoning | 54.1% ±0.3, Mean of 3 runs, range 53.6 to 54.3 | #36 of 63 | $0.0059 | 47.87s |
Overall benchmark score against estimated cost per sample, on a log scale. Upper-left is the sweet spot: high quality at low cost.
62 models on the current benchmark · scores and efficiency pooled across all five tasks at low effort · Qwen3.5 35B-A3B highlighted
Qwen3.5 35B A3B scores are the mean of 3 runs per task at low effort; this model has no separate high-effort tier because its chat template only switches thinking on or off · Methodology
View all Vision Evals →2 quantizations of Qwen3.5 35B-A3B, served with vLLM from the published weights and scored on the same tasks as the hosted models. Expand a row for the GPUs it was measured on.
Rows are ordered by overall score. Every score is the mean of three runs per task under the benchmark protocol (a row marked single run has one), and the gap between quantizations is usually inside the run-to-run range, so read the order as a guide rather than a verdict. Self-hosted rows run with thinking on, the same setting as the hosted frontier models; some hosted open-weight rows still run without thinking, so a self-hosted quant can score above its own hosted API. Quantization still costs a little precision.
Qwen3.5 35B A3B costs $0.150 per 1M input tokens and $1.00 per 1M output tokens.
Pricing updated Oct 13, 2026
Other models worth comparing for similar use cases.
Qwen3.5 35B A3B is released under the Apache 2.0 license, so it's commercially usable out of the box: you can ship it in closed-source, commercial products, including self-hosted deployments, without paying a license fee or open-sourcing your code. The Qwen3.5 35B A3B license is permissive, so no separate commercial license is required.
Apache-2.0 grants an express patent license that terminates if you bring a patent claim over the work, and it disclaims warranties. Validate Qwen3.5 35B A3B on your own data before you depend on it in production.
Read the full Apache 2.0 license ↗This is the straightforward case: a permissive license is the best technical solution and you are free to deploy Qwen3.5 35B A3B commercially without open-sourcing your own code.
Do not hesitate to reach out with questions for your commercial project — our team will help you start solving business problems on the first call. See Roboflow commercial licensing for the models included in each plan.
Talk to salesThis model is released under the Apache License 2.0, a permissive open-source license that allows commercial use, modification, distribution, and patent use.
Yes. Under the terms of the Apache 2.0 license, you can freely use this model for commercial purposes, including in proprietary products. You must retain the copyright notice and disclaimers when redistributing.
License information is provided as a guide and is not legal advice.
Yes. Qwen3.5 35B A3B accepts image input and handles OCR, object counting, identification, visual reasoning, and object detection. On Roboflow's Vision Evals its strongest task is Counting at 62.6% (#33 of 63 at low effort). You can test it on your own image in the demo above.
Not its strength. On Vision Evals, Qwen3.5 35B A3B scores 52.9% mAP@50 on object detection (#34 of 63 at low effort) and 62.6% judge-graded accuracy on object counting. For production counting or precise localization, pairing it with a specialized detector like RF-DETR or your own trained model in a Roboflow Workflow is usually more reliable: detect the objects, then count the detections.
On our benchmark's task mix, Qwen3.5 35B A3B averages $0.0040 per sample at $0.15 per 1M input and $1.00 per 1M output tokens (#23 of 63 on cost), with an average speed of 31.9s per sample across the benchmark. Actual cost depends on your images and prompts.
On the overall Vision Evals ranking, Qwen3.5 35B A3B sits #32 of 63 at 62.5%, just behind Grok 4.7 (63.5%) and just ahead of Gemini 3.5 Flash-Lite (61.7%). See the full side-by-side: Qwen3.5 35B A3B vs Grok 4.7.