Skip to main content
llm.info

Qwen2.5-Max

Alibaba Cloud

Alibaba's large-scale proprietary MoE model (January 2025), pretrained on over 20 trillion tokens and post-trained with SFT and RLHF. Served via the Alibaba Cloud API, Qwen2.5-Max competes with leading frontier models on knowledge, reasoning, and coding benchmarks.

Strengths

  • Large-scale MoE pretrained on 20T+ tokens
  • Post-trained with SFT and RLHF for strong instruction following
  • Competitive with leading frontier models on many benchmarks
  • Backed by Alibaba Cloud's serving infrastructure

Caveats

  • Proprietary - not open weights (unlike the Qwen open models)
  • 32K context smaller than some frontier models
  • API access via Alibaba Cloud only
Vendor
Alibaba Cloud
Context Window
32,768 tokens
License
Proprietary
Release Date
2025-01-28
Modalities
text
BenchmarkScoreMax
livebench_language57.2100.0
livebench_coding64.4100.0
livebench_instruction_following76.6100.0

Capabilities

Vision
Audio
Video
Tool Use

Pricing

Served via the Alibaba Cloud Model Studio API (Qwen-Max); usage-based pricing. Not open weights.

Reviews

Comments