Qwen3 30B A3B
Qwen3 30B A3B is Alibaba's Mixture-of-Experts model with 3.3B activated parameters, hybrid thinking and non-thinking modes, and 119-language support.
Model Overview
Capabilities, design details, and architectural traits
Qwen3 30B A3B - Mixture-of-Experts Model with Hybrid Reasoning Modes
Qwen3 30B A3B is a Mixture-of-Experts large language model from the Qwen3 family designed to balance high performance with efficient inference. It has 30.5 billion total parameters with 3.3 billion active per forward pass.
What Makes It Different: MoE Architecture with Hybrid Thinking Modes
Qwen3 30B A3B's MoE design and dual reasoning modes distinguish it from dense models:
| Distinctive Trait | Why It Matters | How It Works |
|---|---|---|
| 3.3B activated parameters | Efficient inference without performance loss | Uses 8 experts out of 128 total during inference, activating only 3.3B of 30.5B total per token |
| Hybrid reasoning modes | Complex tasks need step-by-step thinking, simple queries need speed | Supports both thinking mode for deep reasoning and non-thinking mode via /no_think for rapid responses |
| 128-expert pool | Broader knowledge coverage with selective activation | Selects 8 optimal experts from 128 pool for each forward pass based on input context |
| 131K token context with YaRN | Long documents and massive datasets require extended context | Native 32K context extended to 131,072 tokens using YaRN rope scaling method |
| Strong-to-weak distillation | Competitive performance from smaller model | Trained via distillation from larger Qwen3 models, maintaining reasoning and coding quality with fewer activated parameters |
Core Application Focus
The model is optimized for conversational AI, code assistance, agentic systems, search, multimedia, and enterprise RAG. It supports over 100 languages for multilingual instruction and translation tasks.
Benchmark Performance
Independent evaluations · Artificial Analysis
Accuracy & Capability Details
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Alibaba | Anthropic | Anthropic |
| Release Date | July 29, 2025 | September 22, 2026 | September 28, 2026 |
| Knowledge Cutoff | Mar 2025 | - | Jun 2026 |
| Context & Limits | |||
| Context Window | 33K | 1M Best Context Window | 1M Best Context Window |
| Pricing (per 1M tokens) | |||
| Input Pricing | $0.20 Best Input Pricing | $4 | $2 |
| Output Pricing | $0.40 Best Output Pricing | $20 | $10 |
| Modalities | |||
| Inputs | text | textimagefile | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 7.5 | 57.6 Best Intelligence Index | 56.0 |
| Coding Index | - | - | - |
| Agentic Index | - | - | - |
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
Long Context Reasoning
Logical reasoning over long context windows.
SciCode Benchmark
Scientific coding and mathematical modeling.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Explore more from Alibaba
Other models by Alibaba
Qwen3.8 Max
Qwen3.8 2.4T A95B
Qwen3.8-Flash-Next
Qwen3.8 27B
Top AI Models
Leading alternatives by intelligence score
