SEAL Precise Instruction Following
Models answer 1,054 private prompts with detailed instructions in nine categories. Human annotators check 12 yes or no criteria and compare responses side by side on a 7-point scale.
One of the four launch SEAL boards. Each model was paired with every other at least 50 times. Scale marked it deprecated in 2025-01 with 19 models, led by o1 at 91.96.
- Level
- Everyday Most adults, without training
- Scored
- Graded by people Human experts grade or compare the outputs.
- Answers
- Open-ended Free-form text, documents, designs or media judged as a whole.
o1 leads with 92.0 among models released by Jan 2025
Best score per entry from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.
The newest model with a score here came out in Jan 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6 Astra
- Claude Opus 5.5
- GPT-6.1 Terra
- GPT-6.1 Sol
- and 16 more
- 92.0 o1
- 87.8
DeepSeek-R1 - 86.6 o1-preview
- 86.6
Gemini 2.0 Flash - 86.0 Claude 3.5 Sonnet
- 85.3 GPT-4o
- 84.8
Llama 3.1-405B - 84.2 Gemini 1.5 Pro (August 27, 2024)
- 83.2 GPT-4 Turbo
- 82.8
Mistral Large 2 - 82.5 GPT-4o
- 82.3 DeepSeek-V3
- 82.1 Llama 3.2 90B
- 81.2 Llama 3-70B
- 80.2 GPT-4o
- 80.1 Claude 3 Opus
- 79.9 Mistral Large
- 79.5 GPT-4 Turbo
- 79.4 Gemini 1.5 Pro
- 1 o1 92.0
- 2 DeepSeek-R1 87.8
- 3 o1-preview 86.6
- 4 Gemini 2.0 Flash Dec 2024 86.6
- 5 Claude 3.5 Sonnet 86.0
- 6 GPT-4o May 2024 85.3
- 7 Llama 3.1-405B 84.8
- 8 Gemini 1.5 Pro (August 27, 2024) 84.2
- 9 GPT-4 Turbo Apr 2024 83.2
- 10 Mistral Large 2 Jul 2024 82.8
- 11 GPT-4o Nov 2024 82.5
- 12 DeepSeek-V3 82.3
- 13 Llama 3.2 90B 82.1
- 14 Llama 3-70B 81.2
- 15 GPT-4o Aug 2024 80.2
- 16 Claude 3 Opus 80.1
- 17 Mistral Large 79.9
- 18 GPT-4 Turbo Nov 2023 79.5
- 19 Gemini 1.5 Pro May 2024 79.4
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | o1 | 92.0 | Dec 2024 | Official leaderboard | |
| 2 | DeepSeek-R1 | 87.8 | Jan 2025 | Official leaderboard | |
| 3 | o1-preview | 86.6 | Sep 2024 | Official leaderboard | |
| 4 | Gemini 2.0 Flash Dec 2024 | 86.6 | Dec 2024 | Official leaderboard | |
| 5 | Claude 3.5 Sonnet | 86.0 | Jun 2024 | Official leaderboard | |
| 6 | GPT-4o May 2024 | 85.3 | May 2024 | Official leaderboard | |
| 7 | Llama 3.1-405B | 84.8 | Jul 2024 | Official leaderboard | |
| 8 | Gemini 1.5 Pro (August 27, 2024) | 84.2 | Official leaderboard | ||
| 9 | GPT-4 Turbo Apr 2024 | 83.2 | Apr 2024 | Official leaderboard | |
| 10 | Mistral Large 2 Jul 2024 | 82.8 | Jul 2024 | Official leaderboard | |
| 11 | GPT-4o Nov 2024 | 82.5 | Nov 2024 | Official leaderboard | |
| 12 | DeepSeek-V3 | 82.3 | Dec 2024 | Official leaderboard | |
| 13 | Llama 3.2 90B | 82.1 | Sep 2024 | Official leaderboard | |
| 14 | Llama 3-70B | 81.2 | Apr 2024 | Official leaderboard | |
| 15 | GPT-4o Aug 2024 | 80.2 | Aug 2024 | Official leaderboard | |
| 16 | Claude 3 Opus | 80.1 | Feb 2024 | Official leaderboard | |
| 17 | Mistral Large | 79.9 | Feb 2024 | Official leaderboard | |
| 18 | GPT-4 Turbo Nov 2023 | 79.5 | Nov 2023 | Official leaderboard | |
| 19 | Gemini 1.5 Pro May 2024 | 79.4 | May 2024 | Official leaderboard |
5 models have held number one since its release in May 2024
Lab claims count from the day the card came out, independent runs from the model’s release.
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
No lab model card reports it yet
We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.
More in Instructions and chat
- IFEval In use Models answer about 500 prompts containing verifiable instructions such as word limits or required keywords. Code checks each instruction for strict and loose accuracy. Gemma 4 31B IT 98.9%
- IFBench In use 58 new out-of-domain verifiable output constraints. Scored by programmatic constraint checks (strict or loose).
Nemotron-Cascade 2 30B-A3B 82.9% - MultiChallenge In use Multi-turn conversations testing instruction retention, inference memory, versioned editing and self-coherence. Graded by LLM judge with rubrics. GPT-6.1 Sol 82%
- Multi-IF In use Models follow verifiable instructions over three-turn conversations in eight languages, extending IFEval. Scored by instruction accuracy at each turn. LFM3-1.7B 89.1%
- AdvancedIF In use Models respond to over 1,600 prompts with complex, multi-turn, and system-level instructions. A judge model checks each response against expert-written rubrics.
MAI-Thinking-1 85% - IFStruct In use Models generate JSON or YAML that must follow a requested schema and phrasing constraints, without constrained decoding. A response passes only if every structural constraint holds. LFM3-1.7B 93.7%