Qwen3.6 35B A3B AI Evaluation | LayerLens - Benchmark Results & Performance

Model

Qwen3.6 35B A3B by Qwen

text+image+video->text

Free credits included. No credit card required.

Qwen3.6-35B-A3B is an open-weight multimodal model from Alibaba Cloud with 35 billion total parameters and 3 billion active parameters per token. It uses a hybrid sparse mixture-of-experts architecture that delivers performance comparable to much larger dense models while maintaining efficiency. The model supports multimodal capabilities including text, image, and video understanding, with strong reasoning abilities for coding and agentic tasks.

Evaluations

across 9 benchmarks

Latency

640ms

Context length

262k

tokens

Cost

$0.16 · $0.97

input · output per 1M tokens

Key takeaways

Qwen3.6 35B A3B is a multimodal model (text+image+video to text) designed for strong reasoning in coding, agentic tasks, and understanding diverse data types.
It uses a hybrid sparse Mixture-of-Experts (MoE) architecture with 35 billion total parameters (3 billion active per token) for efficient performance comparable to larger dense models.
The model demonstrates strong mathematical problem-solving skills, achieving 98.6% accuracy on the MATH-500 benchmark, excelling in Algebra, Geometry, Number Theory, and Prealgebra.
Despite its strengths, the model struggles with specialized visual recognition and domain-specific interpretation, notably in Agriculture (50.0% accuracy) and with nuanced qualitative information in fields like Marketing and Literature.
A key limitation includes misinterpretation of problem wording or diagrams, algebraic rearrangement errors in complex expressions, and difficulty establishing correct vector relationships in certain mathematical problems.

Technical specifications

Context length: 262k tokens
Max output: —
Architecture: MoE
Modalities: text+image+video->text
Parameters: ~35B
Open weights: Yes
Licence: Apache 2.0
Regions: China
Vendor: Qwen
Provider model key: qwen/qwen3.6-35b-a3b