Benchmark Almanac

Higher-difficulty harmful request evaluations (Anthropic)

Claude Sonnet 4.6 99.4% leads since February 2026
Only in system cards

Models answer harder, more obfuscated policy-violating requests. Score is the harmless response rate.

In-house Anthropic evaluation. The tasks are not public.

Made by
Anthropic
Released
February 2026
Kind
Refusing harmful requests
Level
Everyday
Scored
Graded by a model
Answers
Open-ended
Tasks
Kept private
Strongest model tested
Claude Opus 4.7 #30 on the capability index
Last in a system card
April 2026
Last model added
April 2026
4
Models scored in lab cards
4
System cards report it, from 1 lab

Claude Sonnet 4.6 leads with 99.4% among models released by April 2026

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.

The newest model with a score here came out in April 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 99.4 Claude Sonnet 4.6
  2. 99.2 Claude Opus 4.6
  3. 99.1 Claude Mythos Preview
  4. 99.0 Claude Opus 4.7
  1. 1 Claude Sonnet 4.6 99.4%
  2. 2 Claude Opus 4.6 99.2%
  3. 3 Claude Mythos Preview 99.1%
  4. 4 Claude Opus 4.7 99%

Source: Labs’ model cards

#ModelScoreSettingDateSource
1 Claude Sonnet 4.6 99.4% Feb 2026 Lab model card
2 Claude Opus 4.6 99.2% Feb 2026 Lab model card
3 Claude Mythos Preview 99.1% Apr 2026 Lab model card
4 Claude Opus 4.7 99% Apr 2026 Lab model card

2 models have held number one in labs’ own numbers since its release in February 2026

Lab claims count from the day the card came out, independent runs from the model’s release.

Anthropic
Lab claims Passed 85% in Feb 2026, by Claude Opus 4.6
Claude Sonnet 4.6 99.4%

Source: Labs’ model cards

Anthropic’s own number for Claude Opus 4.6 passed 85% in February 2026

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Sonnet 4.6 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Sonnet 4.6

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.

DateDocumentScoreAs the document writes itSection
Apr 16, 2026 Claude Opus 4.7 system card 99 Higher-difficulty violative request evaluations, overall harmless response rate, English only (Table 4.1.3.1.A) Safety
Apr 7, 2026 Claude Mythos Preview system card 99.1 Higher-difficulty violative request evaluations, overall harmless response rate, English only (Table 8.1.2.1.A) Safety
Feb 17, 2026 Claude Sonnet 4.6 system card 99.4 Higher-difficulty violative request evaluations, overall harmless response rate (Table 3.1.3.1.A) Safety
Feb 5, 2026 Claude Opus 4.6 system card 99.2 Higher-difficulty violative request evaluations, overall harmless response rate, English only (Table 3.1.3.1.A) Safety

More in Safety and alignment

All 465 in Safety and alignment
  • FORTRESS In use Paired adversarial and benign national-security misuse prompts. Rubric-graded risk score on adversarial and over-refusal score on benign prompts. Claude Opus 5.5 5.78%
  • StrongREJECT In use Forbidden-prompt jailbreak benchmark with a rubric grader that scores how useful a non-refusal actually is. Inkling 98.6%
  • Gray Swan IPI Arena benchmark In use Indirect prompt-injection attacks from public red-teaming competitions replayed on tool, coding and browser agents. Attack success rate at k attempts. Gemini 4 Carbon 0%
  • DistressBench In use Models reply as a chatbot to 718 simulated users in psychological crisis, in single-turn and multi-turn chats. Three judge models score replies against clinician-written rubrics. Muse Spark 1.2 88.4%
  • Lakera B3 (Backbone Breaker Benchmark) In use Crowdsourced prompt-injection attacks hit backbone LLMs inside agent threat snapshots at three defense levels. The score is attack resistance, 100 minus the board's vulnerability percentage. Claude Opus 4.5 89%
  • IHEval In use Models face conflicts between system, user and tool instructions. Reports a composite score for following the higher-priority instruction. GPT-4o 70%