Meta chemical-agent refusal evaluation
Muse Wonder Full 100% leads since October 2026 Models receive high-risk chemical prompts and the benchmark measures how often they refuse. Covers development, procurement and production stages.
In-house Meta refusal evaluation. The prompts are not public.
- Made by
- Meta
- Kind
- Hazardous knowledge
- Level
- Expert
- Scored
- Graded by a model
- Answers
- Open-ended
- Tasks
- Kept private
- Strongest model tested
- Muse Spark 1.3 #26 on the capability index
- Last in a system card
- October 2026
- Last model added
- October 2026
2 models share the top score of 100%
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
- 1 Muse Wonder Full 100%
- 2 Muse Spark 1.3 100%
- 3 Muse Spark 1.1 99.8%
- 4 Muse Spark 99.4%
Source: Labs’ model cards
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Muse Wonder Full | 100% | Oct 2026 | Lab model card | |
| 2 | Muse Spark 1.3 | 100% | Oct 2026 | Lab model card | |
| 3 | Muse Spark 1.1 | 99.8% | Jul 2026 | Lab model card | |
| 4 | Muse Spark | 99.4% | Jun 2026 | Lab model card |
3 models have held number one in labs’ own numbers since 2023
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Labs’ model cards
Meta’s own number for Muse Spark passed 85% in June 2026
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Labs’ model cards
Labs report it only in system cards
We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark. Some system cards do, below.
4 system cards report it, from 1 lab
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Oct 23, 2026 | Muse Spark 1.3 system card | 100 | Chemical Agents (English) internal refusal rate, Muse Spark 1.3 (API) (Table 2, Figure 6). Multilingual 100.0, adversarial 99.7 | Bio and chem risk |
| Oct 12, 2026 | Muse Wonder Full model card | 100 | Internal Chemistry Refusal, English (CBRN Safety Evaluations table, section 4). Multilingual 100.0, adversarial 99.8 | Bio and chem risk |
| Jul 9, 2026 | Muse Spark 1.1 system card | 99.8 | Refusals: Chemical Agents refusal rate (Table 2, Figure 7) | Bio and chem risk |
| Jun 12, 2026 | Muse Spark system card | 99.4 | Refusals: Chemical Agents refusal rate, deployed system (Table 2, Figure 11) | Bio and chem risk |
More in Biology and chemistry
- GeneBench Pro In use Agents analyze messy simulated genomics and biology datasets to estimate a decision-relevant quantity across 129 problems. Scored automatically against known answers. GPT-5.6 Sol 31.5%
- SUPERChem In use Models solve 500 expert-written chemistry reasoning problems, offered in multimodal and text-only forms. Expert solution paths also allow a reasoning path score.
Tencent Hy4 preview 66.4% - LifeSciBench In use Models complete 750 expert-written life science research tasks, often with attached figures, tables, or sequence files. Graded against expert rubrics as score and pass rate. GPT-Rosalind-6 61.3%
- MedChemBench (OpenAI) In use Models answer chemistry questions from an OpenAI internal set, grouped with OpenAI's life science research evaluations. OpenAI has not published the task format. GPT-Rosalind-6 49.6%
- ProtocolQA Open-Ended In use Models correct errors in published lab protocols using open-ended questions adapted from LAB-Bench ProtocolQA. Scored as pass@1. GPT-6 Cyber 47.2%
- MolecularIQ In use Models answer questions about molecular graphs, such as counting, indexing and constrained generation, and a symbolic checker verifies each answer. gpt-oss-120b 47.5%