# Long-form research on HealthBench Source: [Long-form research on HealthBench]() Collection index: [llms.txt]() Data updated: 2026-05-12T17:00:06.304Z Evaluation of model performance on long-form research tasks using the HealthBench dataset, measured by the overall Score. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Long-form research | | Dataset (including any stated split) | HealthBench | | Metric | Overall Score | | Unit | SCALAR | | Direction | Higher is better | | Results loaded | 14 of 14 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [Overall Score](): Higher is better; unit: SCALAR. ## Recorded results | Method | Overall Score | Result date | Report paper | | --- | --- | --- | --- | | GPT-5 + Search | 59.5 | Not recorded | 6a02b4b3dbc836b87d824fab | | OpenAI Deep Research | 53.8 | Not recorded | 6a02b4b3dbc836b87d824fab | | DR Tulu-8B (RL, 1900 steps) | 50.2 | Not recorded | 6a02b4b3dbc836b87d824fab | | RubricEM-8B (RL, 1400 steps) | 49.3 | Not recorded | 6a02b4b3dbc836b87d824fab | | Gemini 3.1 Pro + Search | 47.5 | Not recorded | 6a02b4b3dbc836b87d824fab | | Tongyi DeepResearch-30B-A3B | 46.2 | Not recorded | 6a02b4b3dbc836b87d824fab | | WebThinker-32B-DPO | 39.4 | Not recorded | 6a02b4b3dbc836b87d824fab | | RubricEM-8B (SFT) | 39 | Not recorded | 6a02b4b3dbc836b87d824fab | | DR Tulu-8B (SFT) | 38.1 | Not recorded | 6a02b4b3dbc836b87d824fab | | WebThinker QwQ-32B | 36.5 | Not recorded | 6a02b4b3dbc836b87d824fab | | WebExplorer-8B | 33.7 | Not recorded | 6a02b4b3dbc836b87d824fab | | Ai2 ScholarQA – Claude Sonnet | 32 | Not recorded | 6a02b4b3dbc836b87d824fab | | Qwen3-8B + Our Search | 24.5 | Not recorded | 6a02b4b3dbc836b87d824fab | | Search-R1-7B | -0.1 | Not recorded | 6a02b4b3dbc836b87d824fab | ### GPT-5 + Search Overall Score: 59.5. Result ID: 6a035c7b47f34ce924e9024f. [Reporting paper]() Recorded evaluation configuration: {"Category":"Closed Deep Research"}. This record does not assert independent reproduction. ### OpenAI Deep Research Overall Score: 53.8. Result ID: 6a035c7b47f34ce924e90250. [Reporting paper]() Recorded evaluation configuration: {"Category":"Closed Deep Research"}. This record does not assert independent reproduction. ### DR Tulu-8B (RL, 1900 steps) Overall Score: 50.2. Result ID: 6a035c7b47f34ce924e90258. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Category":"Open Deep Research Models","Backbone":"8B","Training":"RL","Steps":"1900"}. This record does not assert independent reproduction. ### RubricEM-8B (RL, 1400 steps) Overall Score: 49.3. Result ID: 6a035c7b47f34ce924e9025b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"8B","Training":"RL","Steps":"1400"}. This record does not assert independent reproduction. ### Gemini 3.1 Pro + Search Overall Score: 47.5. Result ID: 6a035c7b47f34ce924e9024e. [Reporting paper]() Recorded evaluation configuration: {"Category":"Closed Deep Research"}. This record does not assert independent reproduction. ### Tongyi DeepResearch-30B-A3B Overall Score: 46.2. Result ID: 6a035c7b47f34ce924e90256. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Open Deep Research Models","Backbone":"30B-A3B"}. This record does not assert independent reproduction. ### WebThinker-32B-DPO Overall Score: 39.4. Result ID: 6a035c7b47f34ce924e90252. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Fixed Pipeline Deep Research","Backbone":"32B","Alignment":"DPO"}. This record does not assert independent reproduction. ### RubricEM-8B (SFT) Overall Score: 39. Result ID: 6a035c7b47f34ce924e9025a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"8B","Training":"SFT"}. This record does not assert independent reproduction. ### DR Tulu-8B (SFT) Overall Score: 38.1. Result ID: 6a035c7b47f34ce924e90257. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Category":"Open Deep Research Models","Backbone":"8B","Training":"SFT"}. This record does not assert independent reproduction. ### WebThinker QwQ-32B Overall Score: 36.5. Result ID: 6a035c7b47f34ce924e90251. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Fixed Pipeline Deep Research","Backbone":"QwQ-32B"}. This record does not assert independent reproduction. ### WebExplorer-8B Overall Score: 33.7. Result ID: 6a035c7b47f34ce924e90255. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Open Deep Research Models","Backbone":"8B"}. This record does not assert independent reproduction. ### Ai2 ScholarQA – Claude Sonnet Overall Score: 32. Result ID: 6a035c7b47f34ce924e90253. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Fixed Pipeline Deep Research","Backbone":"Claude Sonnet"}. This record does not assert independent reproduction. ### Qwen3-8B + Our Search Overall Score: 24.5. Result ID: 6a035c7b47f34ce924e90259. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-8B","Search":"Our Search"}. This record does not assert independent reproduction. ### Search-R1-7B Overall Score: -0.1. Result ID: 6a035c7b47f34ce924e90254. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Open Deep Research Models","Backbone":"R1-7B"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.