Qwen3-Max is an updated release built on the Qwen3 series, offering major improvements in reasoning, instruction following, multilingual support, and long-tail knowledge coverage compared to the January 2025 version. It delivers higher accuracy in math, coding, logic, and science tasks, follows complex instructions in Chinese and English more reliably, reduces hallucinations, and produces higher-quality responses for open-ended Q&A, writing, and conversation. The model supports over 100 languages with stronger translation and commonsense reasoning, and is optimized for retrieval-augmented generation (RAG) and tool calling, though it does not include a dedicated “thinking” mode.
Evaluations
19
across 17 benchmarks
Latency
1.1s
Context length
256k
tokens
Cost
$1.20 · $6
input · output per 1M tokens
Key takeaways
Qwen3 Max is a proprietary MoE (Mixture of Experts) model from Qwen, designed for advanced reasoning, instruction following, multilingual support (over 100 languages), and open-ended Q&A, optimized for Retrieval-Augmented Generation (RAG) and tool calling.
The model demonstrates strong capabilities in mathematical reasoning, achieving 96.2% accuracy on the MATH-500 dataset.
A significant limitation is observed in coding challenges, particularly search algorithms and nuanced code interpretations, evidenced by a very poor 3.75% accuracy on the Terminal-Bench Dataset.
Qwen3 Max features a 256k context length and is updated from the Qwen3 series, aiming to reduce hallucinations and improve response quality, though it lacks a dedicated 'thinking' mode.
The model has 175 parameters and is a text-to-text modality, offering higher accuracy in math, coding, logic, and science tasks, and follows complex instructions more reliably in both Chinese and English.
