BlogLLM Benchmark Scores as Published, Side by Side
Benign request over-refusal (multilingual, Anthropic internal): Results as Published
Every result the catalog holds for Benign request over-refusal (multilingual, Anthropic internal): 6 models from 1 developer, read from 2 pages. Each number is as printed by the company named beside it; nothing is ranked or rescored. Models are listed by developer, newest first.
Area Safety and AlignmentMaintainer Anthropic (internal evaluation)Introduced 2026Data version 2026103001, as of 2026-10-30
Single-turn benign requests that are sensitive in subject matter but appropriate to answer, across 16 policy areas in seven languages; the refusal rate is the over-refusal share.
Benign request refusal rate (all languages)
lower is better
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| Claude Haiku 5.5Anthropic, released 2026-10-07 | 0.82% | Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19 | Setup as reported: Table 4.1.2.A, all tested languages, claude.ai |
| 0.17% | Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19 | Setup as reported: Table 4.1.2.A, all tested languages, API | |
| Claude Sonnet 5.5Anthropic, released 2026-09-28 | 0.2% | Anthropic, Claude Sonnet 5.5 System Cardchecked 2026-10-19; printed on 1 more page | Setup as reported: claude.ai, with the default system prompt; Seven languages (Arabic, English, French, Hindi, Korean, Mandarin Chinese, Russian); 16 policy areas |
| 0.02% | Anthropic, Claude Sonnet 5.5 System Cardchecked 2026-10-19 | Setup as reported: API, without a system prompt; Seven languages (Arabic, English, French, Hindi, Korean, Mandarin Chinese, Russian); 16 policy areas | |
| Claude Opus 5.5Anthropic, released 2026-09-22 | 0% | Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19; printed on 1 more page | Setup as reported: Table 4.1.2.A, all tested languages, API |
| 0.33% | Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19 | Setup as reported: Table 4.1.2.A, all tested languages, claude.ai | |
| Claude Fable 5.1Anthropic, released 2026-09-01 | 0.34% | Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19; printed on 1 more page | Setup as reported: Table 4.1.2.A, all tested languages, claude.ai |
| 0% | Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19 | Setup as reported: Table 4.1.2.A, all tested languages, API | |
| Claude Sonnet 5Anthropic, released 2026-06-30 | 1.54% | Anthropic, Claude Sonnet 5.5 System Cardchecked 2026-10-19 | Setup as reported: claude.ai, with the default system prompt; Seven languages (Arabic, English, French, Hindi, Korean, Mandarin Chinese, Russian); 16 policy areas |
| 0.59% | Anthropic, Claude Sonnet 5.5 System Cardchecked 2026-10-19 | Setup as reported: API, without a system prompt; Seven languages (Arabic, English, French, Hindi, Korean, Mandarin Chinese, Russian); 16 policy areas | |
| Claude Haiku 4.5Anthropic, released 2025-10-15 | 3.05% | Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19 | Setup as reported: Table 4.1.2.A, all tested languages, claude.ai |
| 0.44% | Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19 | Setup as reported: Table 4.1.2.A, all tested languages, API |
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.