Iris-mini 与 Iris-pro 是同级别最强的开放权重搜索智能体

The AllSpark team has released Iris-mini and Iris-pro, two open-source search agents along with a full training recipe. According to the paper, the training data and models also improved performance on tasks they were never trained for, including general tool use and office work. Search agents built on language models research the web on their own. They need to understand the question, decide what to search for, interpret the results, and judge when they've gathered enough evidence for an answer. How much of that the models actually do is debated. On established benchmarks, leading AI systems of this kind mostly use the web to confirm knowledge they already picked up during training . In a new paper, Chinese lab AllSpark describes two search agents of different sizes. Iris-mini has 35 billion parameters, and Iris-pro has 397 billion. Both build on Qwen-series models ( Qwen3.6-35B-A3B and Qwen3.5-397B-A17B ), work with a 256,000-token context window, and deliver the strongest results among open-weight search agents in their respective size class, according to the team. The training pipeline builds tasks backward from the link structure of web pages. Starting from a seed page and its outgoing links, it constructs a graph of terms and relationships. From that graph, it generates a multi-step question whose answer requires chaining several connected steps together. Every term except the final answer gets replaced with a paraphrase, so no clue can be resolved through a simple text search. The agent has to reason, not just look things up. Only questions that a reference model can't solve without tools but can solve with the right sources make it into the dataset. That keeps the tasks both hard and clearly verifiable. A stronger teacher model generates solution paths made up of reasoning, search queries, and results. These paths go through two rounds of filtering. The first checks the full path for correctness, repetition loops, and search depth. The second is a step-by-step review by a judge model whose criteria were derived from the data itself rather than set by hand, according to the paper. After that, the model is improved through reinforcement learning against a live web search. The judge model and result summaries run inside the training cluster, powered by the team's own large Qwen model so training doesn't depend on external services. Supervised fine-tuning and reinforcement learning alternate in a process the authors call "SFT-RL climbing." The hardest solved tasks and the most efficient solution paths from each round feed back into the next training cycle. The team argues that runtime context management on common benchmarks often makes a bigger difference than the reported gaps between systems. During long research sessions, the context can fill up before the agent has resolved all sub-questions. Tricks like discarding the conversation history extend the research artificially but say little about the model's actual quality. To isolate the effect, the team tests every benchmark with and without context management while keeping tools, context limits, and the judge model constant. Results reported only with management turned on can't be cleanly split into what comes from the model and what comes from the scaffolding around it. The Iris scores also come from a single agent, with no helper agents and no extra verification steps at the end. Testing covered BrowseComp, which tests the ability to find rare facts from indirect clues, its Chinese counterpart BrowseComp-ZH, DeepSearchQA, which evaluates the completeness of retrieved evidence, and Humanity's Last Exam , which poses academic questions at expert level. With context management turned on, Iris-mini scores 82.2, 84.8, 86.9, and 52.3 according to the paper. Iris-pro reaches 88.6, 85.1, 92.9, and 56.4. In the smaller class, Iris-mini leads on three of four benchmarks and beats the next-best model, XYZ-Aquila-mini, on BrowseComp by 3.4 points, though it trails on De