CASE-Bench
Models decide whether a request is safe in a given context. The same request appears in two contexts with different human judgments.
- Made by
- Guangzhi Sun and colleagues
- Released
- January 2025
- Kind
- Refusing harmful requests
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Multiple choice
- Tasks
- All public
- Measures
- Capability
- Risk
- Privacy
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Content safety
- Strongest model tested
- Claude 3.5 Sonnet #158 on the capability index
- Last model added
- August 2024
- Scores refreshed
- November 20th, 2026
Claude 3.5 Sonnet leads with 90.1% among models released by August 2024 Claude 3.5 Sonnet leads with 90.1% among models released by August 2024
Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.
The newest model with a score here came out in August 2024, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 90.1 Claude 3.5 Sonnet
- 88.6 DeepSeek-R1 (snapshot unspecified)
- 87.2
Llama 3-70B
- 87.0
Llama 3.1-8B
- 86.4
Qwen2-72B
- 83.1
Mixtral 8x7B
- 82.7
Mistral-7B-Instruct-v0.2
- 82.6 Dolphin-2.9-Llama3-70B
- 82.4
Llama 3-8B
- 82.0 GPT-4o mini
- 81.6
Zephyr 7B Beta
- 78.6 GPT-4o
- 73.4
Starling-LM-7B-alpha
- 60.1 Llama-Guard-3-8B
- 1 Claude 3.5 Sonnet Score 90.1%
- 2 DeepSeek-R1 (snapshot unspecified) Score 88.6%
- 3
Llama 3-70B Prob. 87.2%
- 4
Llama 3.1-8B Score 87%
- 5
Qwen2-72B Binary 86.4%
- 6
Mixtral 8x7B Score 83.1%
- 7
Mistral-7B-Instruct-v0.2 Prob. 82.7%
- 8 Dolphin-2.9-Llama3-70B Binary 82.6%
- 9
Llama 3-8B Prob. 82.4%
- 10 GPT-4o mini Binary 82%
- 11
Zephyr 7B Beta Prob. 81.6%
- 12 GPT-4o Aug 2024, Score 78.6%
- 13
Starling-LM-7B-alpha Prob. 73.4%
- 14 Llama-Guard-3-8B Context 60.1%
Source: The CASE-Bench paper
- 1 Claude 3.5 Sonnet Score 90.1%
- 2
Llama 3-70B Prob. 87.2%
- 3
Llama 3.1-8B Score 87%
- 4
Qwen2-72B Binary 86.4%
- 5
Mixtral 8x7B Score 83.1%
- 6
Mistral-7B-Instruct-v0.2 Prob. 82.7%
- 7
Llama 3-8B Prob. 82.4%
- 8 GPT-4o mini Binary 82%
- 9
Zephyr 7B Beta Prob. 81.6%
- 10 GPT-4o Aug 2024, Score 78.6%
- 11
Starling-LM-7B-alpha Prob. 73.4%
Source: The CASE-Bench paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude 3.5 Sonnet | 90.1% | Score | Jun 2024 | Official leaderboard |
| 2 | DeepSeek-R1 (snapshot unspecified) | 88.6% | Score | Official leaderboard | |
| 3 | | 87.2% | Prob. | Apr 2024 | Official leaderboard |
| 4 | | 87% | Score | Jul 2024 | Official leaderboard |
| 5 | | 86.4% | Binary | Jun 2024 | Official leaderboard |
| 6 | | 83.1% | Score | Dec 2023 | Official leaderboard |
| 7 | | 82.7% | Prob. | Dec 2023 | Official leaderboard |
| 8 | Dolphin-2.9-Llama3-70B | 82.6% | Binary | Official leaderboard | |
| 9 | | 82.4% | Prob. | Apr 2024 | Official leaderboard |
| 10 | GPT-4o mini | 82% | Binary | Jul 2024 | Official leaderboard |
| 11 | | 81.6% | Prob. | Oct 2023 | Official leaderboard |
| 12 | GPT-4o Aug 2024 | 78.6% | Score | Aug 2024 | Official leaderboard |
| 13 | | 73.4% | Prob. | Nov 2023 | Official leaderboard |
| 14 | Llama-Guard-3-8B | 60.1% | Context | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude 3.5 Sonnet | 90.1% | Score | Jun 2024 | Official leaderboard |
| 2 | | 87.2% | Prob. | Apr 2024 | Official leaderboard |
| 3 | | 87% | Score | Jul 2024 | Official leaderboard |
| 4 | | 86.4% | Binary | Jun 2024 | Official leaderboard |
| 5 | | 83.1% | Score | Dec 2023 | Official leaderboard |
| 6 | | 82.7% | Prob. | Dec 2023 | Official leaderboard |
| 7 | | 82.4% | Prob. | Apr 2024 | Official leaderboard |
| 8 | GPT-4o mini | 82% | Binary | Jul 2024 | Official leaderboard |
| 9 | | 81.6% | Prob. | Oct 2023 | Official leaderboard |
| 10 | GPT-4o Aug 2024 | 78.6% | Score | Aug 2024 | Official leaderboard |
| 11 | | 73.4% | Prob. | Nov 2023 | Official leaderboard |
Rankings by part
We read 2 parts from the benchmark’s paper: Safe recall and Unsafe recall. Each ranks models on one part of the test. Llama 3.1-8B leads Safe recall and Starling-LM-7B-alpha leads Unsafe recall. Arrows show how far a model moved from its overall rank among the models we track.
Safe recall
11 models we track have a score here.
- 1
Llama 3.1-8B ↑2 92.2%
- 2 Claude 3.5 Sonnet ↓1 90.7%
- 3
Llama 3-70B ↓1 89.2%
- 4
Qwen2-72B 76.9%
- 5
Zephyr 7B Beta ↑4 73.4%
- 6
Mistral-7B-Instruct-v0.2 72.7%
- 7
Mixtral 8x7B ↓2 71.7%
- 8
Llama 3-8B ↓1 68.9%
- 9 GPT-4o mini ↓1 67.4%
- 10 GPT-4o Aug 2024 58.1%
Unsafe recall
11 models we track have a score here.
- 1
Starling-LM-7B-alpha ↑10 100%
- 2
Zephyr 7B Beta ↑7 98.4%
- 3
Mistral-7B-Instruct-v0.2 ↑3 97.2%
- 4
Llama 3-8B ↑3 96.6%
- 5 GPT-4o Aug 2024 ↑5 94.8%
- 6 GPT-4o mini ↑2 94.4%
- 7
Llama 3.1-8B ↓4 94%
- 8
Qwen2-72B ↓4 93.6%
- 9
Mixtral 8x7B ↓4 92.4%
- 10 Claude 3.5 Sonnet ↓9 90.4%
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: The CASE-Bench paper
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Claude 3.5 Sonnet | Score | 90.1% | |||
| Claude 3.5 Sonnet | Binary | 88.7% | |||
| Prob. | 87.2% | ||||
| Score | 87% | ||||
| Binary | 86.6% | ||||
| Binary | 86.4% | ||||
| Prob. | 84.9% | ||||
| Score | 84.7% | ||||
| Score | 84.4% | ||||
| Score | 83.1% | ||||
| Prob. | 83.1% | ||||
| Prob. | 82.7% | ||||
| Prob. | 82.4% | ||||
| GPT-4o mini | Binary | 82% | |||
| Binary | 81.9% | ||||
| Prob. | 81.6% | ||||
| Prob. | 81.6% | ||||
| GPT-4o mini | Score | 80% | |||
| GPT-4o Aug 2024 | Score | 78.6% | |||
| GPT-4o Aug 2024 | Binary | 77% | |||
| Prob. | 73.4% | ||||
| Binary | 70.4% | ||||
| Binary | 70.3% | ||||
| Score | 70% | ||||
| Score | 66.3% | ||||
| Score | 56.8% | ||||
| Score | 56.8% | ||||
| Binary | 56% | ||||
| Binary | 55.7% | ||||
| Binary | 44.7% |
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: The CASE-Bench paper
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Claude 3.5 Sonnet | Score | 90.1% | |||
| Claude 3.5 Sonnet | Binary | 88.7% | |||
| Prob. | 87.2% | ||||
| Score | 87% | ||||
| Binary | 86.6% | ||||
| Binary | 86.4% | ||||
| Prob. | 84.9% | ||||
| Score | 84.7% | ||||
| Score | 84.4% | ||||
| Score | 83.1% | ||||
| Prob. | 83.1% | ||||
| Prob. | 82.7% | ||||
| Prob. | 82.4% | ||||
| GPT-4o mini | Binary | 82% | |||
| Binary | 81.9% | ||||
| Prob. | 81.6% | ||||
| Prob. | 81.6% | ||||
| GPT-4o mini | Score | 80% | |||
| GPT-4o Aug 2024 | Score | 78.6% | |||
| GPT-4o Aug 2024 | Binary | 77% | |||
| Prob. | 73.4% | ||||
| Binary | 70.4% | ||||
| Binary | 70.3% | ||||
| Score | 70% | ||||
| Score | 66.3% | ||||
| Score | 56.8% | ||||
| Score | 56.8% | ||||
| Binary | 56% | ||||
| Binary | 55.7% | ||||
| Binary | 44.7% |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The CASE-Bench paper
Llama 3-70B on the official leaderboard passed 85% in April 2024
Models released before it came out already scored that high. All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official CASE-Bench leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
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%
- Covert misbehavior (Thinking Machines) In use Agentic tasks that check whether a model misbehaves and conceals it. Scored by the rate of covert misbehavior.
Inkling 2 Mini 8.30%
- Destructive actions (Thinking Machines) In use Agentic tasks that check whether a model takes destructive actions it was not authorized to take. Scored by the rate of destructive actions.
Inkling 2 Small 4.70%
- Misbehavior (Thinking Machines) In use Agentic tasks that check whether a model fabricates results, ignores instructions or hides what it did when a task is hard. Scored by the rate of misbehavior.
Inkling 2 Mini 9.80%
Data audit
All 8 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
- Pass Independent runs Scores from the benchmark’s paper (14 entries).
- Pass Models scored 11 models we track, plus 3 other entries on its leaderboard.
- Note Newest score The newest came 2 years ago, in August 2024, for GPT-4o. That’s expected for a saturated test.
- Note Frontier models None of the top 10 has a score, which is fine for a retired test. The strongest tested is Claude 3.5 Sonnet (#158).
- Note Headroom Saturated. The best is 90.1%, past 85% since April 2024.
Leaderboard import 3 pass
- Pass Importer We read the board on arxiv.org: 30 rows for models we track and 3 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.