BlogLLM Benchmark Scores as Published, Side by Side
ResearchRubrics: Results as Published
Every result the catalog holds for ResearchRubrics: 5 models from 4 developers, read from 1 page. Each number is as printed by the company named beside it; nothing is ranked or rescored. Models are listed by developer, newest first.
Area Web Search and ResearchMaintainer ResearchRubrics authors (lead author Manasi Sharma); the maintaining organisation is not named on the arXiv pageIntroduced 2025Data version 2026103001, as of 2026-10-30
ResearchRubrics tests whether a deep research agent can answer open-ended research prompts that are judged against expert-written rubrics. The benchmark contains realistic, domain-diverse prompts and more than 2,500 fine-grained expert-written rubrics. The reported result is average rubric compliance: the paper's abstract reports that the leading deep research systems reach under 68% average compliance.
ResearchRubrics
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPT-5.6 SolOpenAI, released 2026-07-09 | 73.8% | Moonshot AI, Kimi K3 model cardchecked 2026-10-15 | Effort: max |
| GPT-5.5OpenAI, released 2026-04-23 | 64% | Moonshot AI, Kimi K3 model cardchecked 2026-10-15 | Effort: xhigh |
| Claude Opus 4.8Anthropic, released 2026-05-28 | 73.5% | Moonshot AI, Kimi K3 model cardchecked 2026-10-15 | Effort: max |
| Kimi K3Moonshot AI, released 2026-07-16 | 76.2% | Moonshot AI, Kimi K3 model cardchecked 2026-10-15 | Effort: max |
| GLM-5.2Z.ai, released 2026-06-16 | 71.1% | Moonshot AI, Kimi K3 model cardchecked 2026-10-15 | Effort: max |
Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.