Skip to content

AI-Needle

A long-context retrieval benchmark that measures whether a model can recover relevant information embedded deep inside very long contexts.

4 models

Higher is better.

  1. 1Claude Opus 4.5Anthropic74.0
  2. 2Qwen3.5 397BAlibaba68.7
  3. 3Qwen3.6 PlusAlibaba68.3
  4. 4GLM-5Z.AI63.3
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Needle-in-a-haystack recall
Tasks
Long-context retrieval
Authors
Qwen (2026)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.