Hasty Briefsbeta

双语

Qwen3.8 2.4T A95B: Intelligence, Performance and Price Analysis

a month ago
  • Qwen3.8 2.4T A95B scores 58 on the Artificial Analysis Intelligence Index, well above the median of 27 for comparable open-weight models.
  • It is expensive: $2.00 per 1M input tokens and $6.00 per 1M output tokens, significantly above median prices.
  • The model has a large 984k token context window and supports text input and output only.
  • Output speed is slow at 49 tokens per second (median 69), with a somewhat high time to first token of 2.63 seconds.
  • It is a reasoning model using chain-of-thought, with 2.4 trillion total parameters and 95 billion active parameters.
  • Pricing evaluation cost $1,671.89 on the Intelligence Index, and the model generated 140M output tokens, indicating verbosity.
  • Released on August 12, 2026, by Alibaba under the Qwen3.8-Max License, with open weights available via API and self-hosting.