TAU2
Measured May 14, 2026Source
Score
0.89
Models9 2 
Qwen3.5 35B A3B (Reasoning)
阿里巴巴阿里巴巴China
A 35B parameter reasoning model from the Qwen3.5 series, utilizing a Mixture-of-Experts architecture with 3B activated parameters for efficient inference. Optimized for complex reasoning tasks while maintaining strong performance in coding and long-context understanding.
ReasoningCodingFastLong contextMultimodal
Input / 1M tokens
$0.25Output / 1M tokens
$2.00Output tokens/s
122.98First-token seconds
1.09sSupported plans
7Benchmark history
Evaluations
Terminalbench Hard
Measured May 14, 2026Source
Score
0.27
Lcr
Measured May 14, 2026Source
Score
0.63
Ifbench
Measured May 14, 2026Source
Score
0.73
Scicode
Measured May 14, 2026Source
Score
0.38
Hle
Measured May 14, 2026Source
Score
0.2
Gpqa
Measured May 14, 2026Source
Score
0.85
Artificial Analysis Coding Index
Measured May 14, 2026Source
Score
30.3
Artificial Analysis Intelligence Index
Measured May 14, 2026Source
Score
37.1
Plan availability
Products and plans that support this model
Apertis Coding Plan
Apertis Coding Plan
Apertis Coding Plan is a subscription-based AI coding service providing unified access to 30+ AI models (GPT-5.4, Claude Opus 4.6, Gemini 3.1 Pro, and more) through a single API key. Designed for developers using coding agents like Claude Code, Cursor, Cline, and OpenCode, it offers predictable monthly pricing, free prompt caching, auto-failover, and quota-based billing across OpenAI, Anthropic, Google, and other providers.
Astron Coding Plan
Astron Coding Plan
Astron Coding Plan is a monthly AI coding subscription service by XFYun MaaS for developers, providing access to models like Spark X2, Qwen3.5-35B-A3B, DeepSeek-V3.2, GLM-5, MiniMax-M2.5, and Kimi-K2.5 via OpenAI and Anthropic protocols, compatible with coding tools such as OpenClaw, Claude Code, Cursor, and OpenCode.
Discussion
Thinking... Make sure you are connected to GitHub server