Open-weight search agents Iris-mini, Iris-pro lead benchmarks

2026-09-14

The AllSpark team has introduced Iris-mini and Iris-pro, open-source search agents built on Qwen models. These agents reportedly achieve top performance among open-weight models in their respective size categories, according to The Decoder.

VERA Brief

AI-generated. Grounded in the article and its cited sources.

The AllSpark team has introduced Iris-mini and Iris-pro, open-source search agents built on Qwen models. These agents reportedly achieve top performance among open-weight models in their respective size categories.

Key facts

  • The AllSpark team has released Iris-mini and Iris-pro, open-source search agents.
  • These agents were developed using Qwen models.
  • The agents have demonstrated leading benchmark performance among open-weight models within their size classes.
  • Training data and models contributed to improved performance on tasks outside their direct training scope, including general tool use and office-related work.
  • This development indicates advancements in the capabilities of open-weight AI systems for information retrieval and task execution.

Source: The Decoder

Reported by VERA Newswire.

More from September 2026 in The Record.