ChengRang

Xiaohongshu Iris (AllSpark)

AI Platforms Open Source

Search-agent model family open-sourced by Xiaohongshu's AllSpark team on Sep 14, 2026: Iris-mini (35B, built on Qwen3.6-35B-A3B) and Iris-pro (397B, built on Qwen3.5-397B-A17B), both with 256K context; leads open models of comparable size across BrowseComp, BrowseComp-ZH, DeepSearchQA and HLE, with Iris-mini's 82.2 on BrowseComp approaching trillion-parameter models; trained with an SFT-to-RL climbing recipe, in-cluster judging and summarization, and resumable ultra-long trajectories; weights, paper and model cards are public

LLMSearch AgentOpen SourceXiaohongshu
Visit Xiaohongshu Iris (AllSpark)

Disclaimer: Review content represents our editorial team's views and experience, not commercial recommendation or investment advice. Product info and pricing may change; refer to official sources.

Overview

Iris is a pair of search-agent models open-sourced by Xiaohongshu's AllSpark team on September 14, 2026: the 35B Iris-mini and the 397B Iris-pro, built on Qwen3.6-35B-A3B and Qwen3.5-397B-A17B respectively, both with 256K context. The team focused on one job. Training data is generated backwards from the hyperlink structure of the web: seed pages spawn an entity graph and multi-hop questions, then every entity except the answer is rewritten as a descriptive reference to erase the clues a model could simply copy. A sample survives only if the model cannot answer with the web closed and can answer correctly and uniquely with it open.

Evaluation covers BrowseComp, BrowseComp-ZH, DeepSearchQA and HLE. Iris-pro scores 88.6, 85.1, 92.9 and 56.4; Iris-mini scores 82.2, 84.8, 86.9 and 52.3, with its 82.2 on BrowseComp approaching models many times its size. Both sit at the front of open models in their parameter class.

Key Features

Use Cases

Pros

Pricing

Weights are released as open source and can be self-hosted; inference cost depends on the version and the hosting platform. The two sizes map to different VRAM and throughput budgets.

Summary

Iris suits teams that treat search as a core capability: deep research Q&A, agents that need dependable retrieval, or Chinese-language scenarios where answers must be verifiable. The 35B version is friendlier to local deployment while the 397B version targets the ceiling. If you only need occasional web answers, a general model's built-in browsing is usually enough and a dedicated search model is not worth the extra moving part.

Version History

Category
AI Platforms
Pricing
Open Source
Tags
LLM · Search Agent · Open Source
Website

Related Tools