Qwen: Qwen3.5-35B-A3B

Image input Text input Video input Text output
Author's Description

The Qwen3.5 Series 35B-A3B is a native vision-language model designed with a hybrid architecture that integrates linear attention mechanisms and a sparse mixture-of-experts model, achieving higher inference efficiency. Its overall performance is comparable to that of the Qwen3.5-27B.

Key Specifications
Cost
$$$$$
Context
262K
Parameters
35B
Released
Feb 25, 2026
Speed
Ability
Reliability
Supported Parameters

This model supports the following parameters:

Presence Penalty Tools Logprobs Temperature Seed Reasoning Include Reasoning Response Format Structured Outputs Top Logprobs Top P Tool Choice Max Tokens
Features

This model supports the following features:

Reasoning Structured Outputs Tools Response Format
Performance Summary

The Qwen3.5-35B-A3B model, a native vision-language model with a hybrid architecture, demonstrates exceptional speed and competitive pricing. It consistently ranks among the fastest models across 8 benchmarks and offers among the most competitive pricing across 4 benchmarks. The model exhibits strong reliability with an 88% success rate, indicating consistent operational stability. In terms of benchmark performance, Qwen3.5-35B-A3B shows notable strengths in specific areas. It achieves a high 98.0% accuracy in Hallucinations (Baseline), placing it in the 64th percentile, and demonstrates strong Instruction Following with 74.7% accuracy (82nd percentile). Its Email Classification accuracy is also impressive at 99.0% (74th percentile). However, a significant weakness is apparent in its performance on General Knowledge, Coding, Reasoning, Ethics, and Mathematics benchmarks, where it recorded 0.0% accuracy across all these categories. This suggests the model, while efficient and reliable for certain tasks, may not be suitable for applications requiring broad factual recall, complex problem-solving, or ethical judgment. Its hybrid architecture and integration of linear attention and sparse mixture-of-experts contribute to its high inference efficiency, making it a strong contender for tasks where speed and cost-effectiveness are paramount, provided the task aligns with its demonstrated strengths.

Model Pricing

Current Pricing

Feature Price (per 1M tokens)
Prompt $0.25
Completion $2

Price History

Available Endpoints
Provider Endpoint Name Context Length Pricing (Input) Pricing (Output)
Alibaba
Alibaba | qwen/qwen3.5-35b-a3b-20260224 262K $0.25 / 1M tokens $2 / 1M tokens
Benchmark Results
Benchmark Category Reasoning Strategy Free Executions Accuracy Cost Duration
Other Models by qwen