Agentic code behavior scores (Anthropic)
Agents work through over one hundred realistic coding scenarios. Graders score instruction following, safety, verification, efficiency, adaptability and honesty out of 10.
In-house Anthropic evaluation. Cards report six dimension scores with no overall score.
- Made by
- Anthropic
- Released
- February 2026
- Kind
- Deception and alignment
- Level
- Expert
- Scored
- Graded by a model
- Answers
- Agent task
- Tasks
- Kept private
- Last in a system card
- April 2026
- Last model added
- April 2026
4
System cards report it, from 1 lab
No scores here yet
No lab, leaderboard or Epoch AI run has reported a score for Agentic code behavior scores (Anthropic) yet.
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 |
|---|---|---|---|---|
| Apr 16, 2026 | Claude Opus 4.7 system card | Agentic Code Behavior Scores out of 10, no system prompt, no overall: instruction following 9.0, safety 9.6, verification 9.0, efficiency 7.7, adaptability 9.8, honesty 10.0. With system prompt 8.6, 9.9, 9.3, 8.1, 9.7, 10.0 (Figure 6.2.2.2.B) | Alignment | |
| Apr 7, 2026 | Claude Mythos Preview system card | Agentic Code Behavior Scores out of 10, no system prompt: instruction following 8.9, safety 9.3, verification 9.2, efficiency 7.6, adaptability 9.8, honesty 10.0, no overall. With system prompt 8.9, 10.0, 9.3, 7.7, 9.8, 10.0 (Figure 4.2.2.2.B) | Alignment | |
| Feb 17, 2026 | Claude Sonnet 4.6 system card | Agentic Code Behavior Scores out of 10, no prompt: instruction following 8.4, safety 8.9, verification 8.7, efficiency 7.1, adaptability 9.5, honesty 9.9, no overall (Figure 4.3.2.B) | Alignment | |
| Feb 5, 2026 | Claude Opus 4.6 system card | Agentic Code Behavior Rates, six dimensions scored out of 10 (instruction following, safety, verification, efficiency, adaptability, honesty), with and without system prompt, no overall, chart without printed values (Figure 6.2.3.2.B) | Alignment |
More 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%