# Deep Research on HealthBench Source: [Deep Research on HealthBench]() Collection index: [llms.txt]() Data updated: 2026-06-15T06:34:27.326Z Evaluation of model performance on the Deep Research task using the HealthBench dataset, measured by the Score metric. 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 | Deep Research | | Dataset (including any stated split) | HealthBench | | Metric | Score | | Unit | SCALAR | | Direction | Higher is better | | Results loaded | 38 of 38 | | 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 - [Score](): Higher is better; unit: SCALAR. ## Recorded results | Method | Score | Result date | Report paper | | --- | --- | --- | --- | | GPT-5 + Search | 59.5 | Not recorded | 69253885d6b0dd334292f936 | | GPT-5 + Search | 59.5 | Not recorded | 6a2f870cd33dfb23e783bf25 | | HOTE-8B | 54.4 | Not recorded | 6a2f870cd33dfb23e783bf25 | | OpenAI Deep Research | 53.8 | Not recorded | 69253885d6b0dd334292f936 | | OpenAI Deep Research | 53.8 | Not recorded | 6a2f870cd33dfb23e783bf25 | | DR Tulu-8B (RL) | 52.8 | Not recorded | 69253885d6b0dd334292f936 | | Dr. Zero-8B | 52.1 | Not recorded | 6a2f870cd33dfb23e783bf25 | | GSPO | 51 | Not recorded | 6a2f870cd33dfb23e783bf25 | | REINFORCE++ | 50.8 | Not recorded | 6a2f870cd33dfb23e783bf25 | | DR Tulu-8B-RL | 50.2 | Not recorded | 6a2f870cd33dfb23e783bf25 | | SPICE-8B | 50.2 | Not recorded | 6a2f870cd33dfb23e783bf25 | | GRPO | 49.6 | Not recorded | 6a2f870cd33dfb23e783bf25 | | Tongyi DeepResearch-30B-A3B | 46.2 | Not recorded | 69253885d6b0dd334292f936 | | Tongyi DeepResearch-30B-A3B | 46.2 | Not recorded | 6a2f870cd33dfb23e783bf25 | | WebThinker-32B-DPO (report) | 39.4 | Not recorded | 69253885d6b0dd334292f936 | | WebThinker-32B-DPO (report) | 39.4 | Not recorded | 6a2f870cd33dfb23e783bf25 | | DR Tulu-8B (SFT) | 38.1 | Not recorded | 69253885d6b0dd334292f936 | | DR Tulu-8B-SFT | 38.1 | Not recorded | 6a2f870cd33dfb23e783bf25 | | Gemini 3 Pro + Search | 38 | Not recorded | 69253885d6b0dd334292f936 | | Gemini 3 Pro + Search | 38 | Not recorded | 6a2f870cd33dfb23e783bf25 | | WebThinker QwQ-32B (report) | 36.5 | Not recorded | 69253885d6b0dd334292f936 | | WebThinker QwQ-32B (report) | 36.5 | Not recorded | 6a2f870cd33dfb23e783bf25 | | WebExplorer-8B | 33.7 | Not recorded | 69253885d6b0dd334292f936 | | WebExplorer-8B | 33.7 | Not recorded | 6a2f870cd33dfb23e783bf25 | | Ai2 ScholarQA - Claude Sonnet | 32 | Not recorded | 69253885d6b0dd334292f936 | | Ai2 ScholarQA-Claude Sonnet (report) | 32 | Not recorded | 6a2f870cd33dfb23e783bf25 | | GPT-5 + Our Search | 31.1 | Not recorded | 69253885d6b0dd334292f936 | | QwQ-32B | 24.5 | Not recorded | 69253885d6b0dd334292f936 | | Qwen3-235B-A22B | 21.3 | Not recorded | 6a2f870cd33dfb23e783bf25 | | Qwen3-8B | 16.5 | Not recorded | 69253885d6b0dd334292f936 | | WebThinker-32B-DPO | 11.1 | Not recorded | 69253885d6b0dd334292f936 | | WebThinker-32B-DPO | 11.1 | Not recorded | 6a2f870cd33dfb23e783bf25 | | Qwen3-8B + Our Search | 5.9 | Not recorded | 69253885d6b0dd334292f936 | | Qwen3-8B | 5.9 | Not recorded | 6a2f870cd33dfb23e783bf25 | | Search-R1-7B | -0.1 | Not recorded | 69253885d6b0dd334292f936 | | Search-R1-7B | -0.1 | Not recorded | 6a2f870cd33dfb23e783bf25 | | ASearcher-Web-7B | -13 | Not recorded | 69253885d6b0dd334292f936 | | ASearcher-Web-7B | -13 | Not recorded | 6a2f870cd33dfb23e783bf25 | ### GPT-5 + Search Score: 59.5. Result ID: 6a0ab2e047f34ce924f0286f. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"Closed Deep Research","Evaluation Protocol":"100-sample subset"}. This record does not assert independent reproduction. ### GPT-5 + Search Score: 59.5. Result ID: 6a2f9cdefbc373592c2b38a9. [Reporting paper]() Recorded evaluation configuration: {"Category":"Closed Deep Research"}. This record does not assert independent reproduction. ### HOTE-8B Score: 54.4. Result ID: 6a2f9cdefbc373592c2b38bc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Open Evolving Deep Research"}. This record does not assert independent reproduction. ### OpenAI Deep Research Score: 53.8. Result ID: 6a0ab2e047f34ce924f02870. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"Closed Deep Research","Evaluation Protocol":"100-sample subset"}. This record does not assert independent reproduction. ### OpenAI Deep Research Score: 53.8. Result ID: 6a2f9cdefbc373592c2b38aa. [Reporting paper]() Recorded evaluation configuration: {"Category":"Closed Deep Research"}. This record does not assert independent reproduction. ### DR Tulu-8B (RL) Score: 52.8. Result ID: 6a0ab2e047f34ce924f0287d. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model Category":"Open Deep Research (Ours)","Training Strategy":"RL"}. This record does not assert independent reproduction. ### Dr. Zero-8B Score: 52.1. Result ID: 6a2f9cdefbc373592c2b38bb. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Category":"Evolving Methods"}. This record does not assert independent reproduction. ### GSPO Score: 51. Result ID: 6a2f9cdefbc373592c2b38b8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"RL Methods"}. This record does not assert independent reproduction. ### REINFORCE++ Score: 50.8. Result ID: 6a2f9cdefbc373592c2b38b9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"RL Methods"}. This record does not assert independent reproduction. ### DR Tulu-8B-RL Score: 50.2. Result ID: 6a2f9cdefbc373592c2b38b6. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Category":"Open Deep Research"}. This record does not assert independent reproduction. ### SPICE-8B Score: 50.2. Result ID: 6a2f9cdefbc373592c2b38ba. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Evolving Methods"}. This record does not assert independent reproduction. ### GRPO Score: 49.6. Result ID: 6a2f9cdefbc373592c2b38b7. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Category":"RL Methods"}. This record does not assert independent reproduction. ### Tongyi DeepResearch-30B-A3B Score: 46.2. Result ID: 6a0ab2e047f34ce924f02877. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open Deep Research Models"}. This record does not assert independent reproduction. ### Tongyi DeepResearch-30B-A3B Score: 46.2. Result ID: 6a2f9cdefbc373592c2b38b1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Open Deep Research Models"}. This record does not assert independent reproduction. ### WebThinker-32B-DPO (report) Score: 39.4. Result ID: 6a0ab2e047f34ce924f02879. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Fixed Pipeline Deep Research"}. This record does not assert independent reproduction. ### WebThinker-32B-DPO (report) Score: 39.4. Result ID: 6a2f9cdefbc373592c2b38b3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Fixed Pipeline Deep Research"}. This record does not assert independent reproduction. ### DR Tulu-8B (SFT) Score: 38.1. Result ID: 6a0ab2e047f34ce924f0287c. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model Category":"Open Deep Research (Ours)","Training Strategy":"SFT"}. This record does not assert independent reproduction. ### DR Tulu-8B-SFT Score: 38.1. Result ID: 6a2f9cdefbc373592c2b38b5. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Category":"Open Deep Research"}. This record does not assert independent reproduction. ### Gemini 3 Pro + Search Score: 38. Result ID: 6a0ab2e047f34ce924f0286d. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"Closed Deep Research"}. This record does not assert independent reproduction. ### Gemini 3 Pro + Search Score: 38. Result ID: 6a2f9cdefbc373592c2b38a8. [Reporting paper]() Recorded evaluation configuration: {"Category":"Closed Deep Research"}. This record does not assert independent reproduction. ### WebThinker QwQ-32B (report) Score: 36.5. Result ID: 6a0ab2e047f34ce924f02878. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Fixed Pipeline Deep Research"}. This record does not assert independent reproduction. ### WebThinker QwQ-32B (report) Score: 36.5. Result ID: 6a2f9cdefbc373592c2b38b2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Fixed Pipeline Deep Research"}. This record does not assert independent reproduction. ### WebExplorer-8B Score: 33.7. Result ID: 6a0ab2e047f34ce924f02875. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open Deep Research Models"}. This record does not assert independent reproduction. ### WebExplorer-8B Score: 33.7. Result ID: 6a2f9cdefbc373592c2b38af. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Open Deep Research Models"}. This record does not assert independent reproduction. ### Ai2 ScholarQA - Claude Sonnet Score: 32. Result ID: 6a0ab2e047f34ce924f0287a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Fixed Pipeline Deep Research","Evaluation Protocol":"100-sample subset"}. This record does not assert independent reproduction. ### Ai2 ScholarQA-Claude Sonnet (report) Score: 32. Result ID: 6a2f9cdefbc373592c2b38b4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Fixed Pipeline Deep Research"}. This record does not assert independent reproduction. ### GPT-5 + Our Search Score: 31.1. Result ID: 6a0ab2e047f34ce924f0286e. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"Closed Deep Research"}. This record does not assert independent reproduction. ### QwQ-32B Score: 24.5. Result ID: 6a0ab2e047f34ce924f02872. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"Naive RAG"}. This record does not assert independent reproduction. ### Qwen3-235B-A22B Score: 21.3. Result ID: 6a2f9cdefbc373592c2b38ac. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Open Deep Research Models"}. This record does not assert independent reproduction. ### Qwen3-8B Score: 16.5. Result ID: 6a0ab2e047f34ce924f02871. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Naive RAG"}. This record does not assert independent reproduction. ### WebThinker-32B-DPO Score: 11.1. Result ID: 6a0ab2e047f34ce924f02876. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open Deep Research Models"}. This record does not assert independent reproduction. ### WebThinker-32B-DPO Score: 11.1. Result ID: 6a2f9cdefbc373592c2b38b0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Open Deep Research Models"}. This record does not assert independent reproduction. ### Qwen3-8B + Our Search Score: 5.9. Result ID: 6a0ab2e047f34ce924f0287b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open Deep Research (Ours)"}. This record does not assert independent reproduction. ### Qwen3-8B Score: 5.9. Result ID: 6a2f9cdefbc373592c2b38ab. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Open Deep Research Models"}. This record does not assert independent reproduction. ### Search-R1-7B Score: -0.1. Result ID: 6a0ab2e047f34ce924f02873. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open Deep Research Models"}. This record does not assert independent reproduction. ### Search-R1-7B Score: -0.1. Result ID: 6a2f9cdefbc373592c2b38ad. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Open Deep Research Models"}. This record does not assert independent reproduction. ### ASearcher-Web-7B Score: -13. Result ID: 6a0ab2e047f34ce924f02874. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open Deep Research Models"}. This record does not assert independent reproduction. ### ASearcher-Web-7B Score: -13. Result ID: 6a2f9cdefbc373592c2b38ae. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Open Deep Research Models"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.