Qwen-Max, based on Qwen2.5, is a large-scale MoE language model optimized for high-performance inference on complex multi-step tasks, trained on over 20 trillion tokens with supervised fine-tuning and RLHF. It is intended for developers and enterprises seeking efficient, accurate natural language understanding and generation at scale.

Provider: QwenProprietaryNo API
ELO: 1366
More Benchmarks
Context: 32.8K

LLM Specifications

Context Length:32.8K
Max Output:8.2K

Pricing

Input Cost:$1.60 / 1M tokens
Output Cost:$6.40 / 1M tokens

Performance Metrics

LiveBench LiveBench Global Avg:51.93

Supported Formats

Text