Back to Explore
Qwen: Qwen3.5-35B-A3B
Alibaba·Released 2026-02-25
Open Source35B262K ctxApache 2.0Multimodal
About
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...
Related Models
Pricing
Input
$0.14
per 1M tokens
Output
$1.00
per 1M tokens
Blended
$0.35
per 1M tokens
Cheaper than 58% of models. Median price is $0.54/1M tokens.
Cost Calculator
Tokens per day1M
100K100M
Daily
$0.35
Monthly
$10.65
vs. Similar Models
Claude 4.5 Sonnet (Non-reasoning)Q:0.0
$6.00+1590%
Qwen3.5 27B (Non-reasoning)Q:0.0
$0.88+146%
Qwen3.6 27B (Non-reasoning)Q:0.0
$1.35+280%
Gemma 4 31B (Reasoning)Q:+0.1
$0.20-42%
Performance
163
tokens/sec
Faster than 79% of models
1.15
seconds
Faster than 47% of models
13.40
seconds
Faster than 35% of models
Market Median
94 tok/s
73% faster
Median TTFT
1.11s
3% slower
Throughput/Dollar
460
tok/s per $/1M
Speed Comparison
GPT-5.4 nano (Non-Reasoning)
163 tok/s-0%
GPT-5.4 mini (Non-Reasoning)
162 tok/s-0%
OpenAI: GPT-5.4 Nano
162 tok/s-1%
Context Window
262K
tokens
Larger than 62% of models
Max Output
82K
tokens
31% of context
Benchmarks
MMLU-ProNot evaluated
GPQA Diamond
84.5%
HLE
19.7%
LiveCodeBenchNot evaluated
SciCode
37.7%
TerminalBench Hard
26.5%
MATH-500Not evaluated
AIMENot evaluated
AIME 2025Not evaluated
IFBench
72.5%
Long Context Recall
62.7%
Tau2
89.2%
Market AverageTop Score
Open Source
apache-2.035BGGUF / GPTQ / AWQ
Downloads
2.0M
Likes
1.4K
VRAM (FP16)
48-80 GB
GPU
A100 80GB