Scale Labs SEAL Instruction Following (SEAL Instruction Following)
We show this table for reference; we do not rank on it.
Scale’s SEAL board for precise instruction following on 1,054 private prompts with detailed, specific commands. BenchLM shows it as display-only reference data, excluded from rankings. Scale marked this board deprecated as of January 2025 and no longer adds models.
Scale score on SEAL Instruction Following — November 12, 2026 snapshot
We mirror the published scale score view for SEAL Instruction Following. o1 (December 2024) leads the public snapshot at 92.0%, followed by DeepSeek-R1 (87.8%) and o1-preview (86.6%). We do not use these results to rank models overall.
o1 (December 2024)
OpenAI
DeepSeek-R1
DeepSeek
o1-preview
OpenAI
19 modelsInstruction FollowingCurrentDisplay onlyUpdated November 12, 2026 snapshot
Scale score table (19 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | o1 (December 2024)OpenAI | 92.0% | Not reported | Unknown |
| 2 | DeepSeek-R1DeepSeek | 87.8% | Not reported | Open |
| 3 | o1-previewOpenAI | 86.6% | Not reported | Closed |
| 4 | Gemini 2.0 Flash Experimental (December 2024)Google | 86.6% | Not reported | Unknown |
| 5 | Claude 3.5 Sonnet (June 2024)Anthropic | 86.0% | Not reported | Unknown |
| 6 | GPT-4o (May 2024)OpenAI | 85.3% | Not reported | Unknown |
| 7 | Llama 3.1 405B InstructMeta | 84.8% | Not reported | Unknown |
| 8 | Gemini 1.5 Pro (August 27, 2024)Google | 84.2% | Not reported | Unknown |
| 9 | GPT-4 Turbo PreviewOpenAI | 83.2% | Not reported | Unknown |
| 10 | Mistral Large 2Mistral | 82.8% | Not reported | Closed |
| 11 | GPT-4o (November 2024)OpenAI | 82.5% | Not reported | Unknown |
| 12 | DeepSeek V3DeepSeek | 82.3% | Not reported | Open |
| 13 | Llama 3.2 90B Vision InstructMeta | 82.1% | Not reported | Unknown |
| 14 | Llama 3 70B InstructMeta | 81.2% | Not reported | Unknown |
| 15 | GPT-4o (August 2024)OpenAI | 80.2% | Not reported | Unknown |
| 16 | Claude 3 OpusAnthropic | 80.1% | Not reported | Closed |
| 17 | Mistral LargeMistral | 79.9% | Not reported | Unknown |
| 18 | GPT-4 (November 2024)OpenAI | 79.5% | Not reported | Unknown |
| 19 | Gemini 1.5 Pro (May 2024)Google | 79.4% | Not reported | Unknown |
How to read this leaderboard
Compare the published configurations as complete evaluation systems. The source can combine a base model, agent scaffold, tools, budget, and inference setting in each result.
Operator receipt: 19 sourced rows are currently displayable on this page; the leading published row is o1 (December 2024) at 92.0%.
Honest limit: This Scale table is display-only context, not benchmark provenance or a weighted model-only comparison. Scale marked this board deprecated as of January 2025 and no longer adds models.
How BenchLM shows SEAL Instruction Following
BenchLM mirrors 19 published rows from Scale Labs’ public SEAL Instruction Following leaderboard, captured on November 12, 2026 snapshot.
The table is display only. It is useful context for a published agent or model configuration, but it does not enter BenchLM’s overall or category rankings. Scale marked this board deprecated as of January 2025 and no longer adds models.
Snapshot
The published SEAL Instruction Following snapshot places o1 (December 2024) first at 92.0%. The third row is 5.4 points behind. The broader top-10 range is 9.1 points, so many of the published results sit in a relatively narrow band.
19 models have been evaluated on SEAL Instruction Following. The benchmark falls in the Instruction Following category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. SEAL Instruction Following is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About SEAL Instruction Following
Year
2026
Tasks
1,054 prompts
Format
Expert-rated score
Difficulty
External agent and model evaluation
Unlike IFEval, the prompts are not limited to instructions a program can check; expert raters judge whether each reply follows every instruction. BenchLM mirrors 19 published rows captured on November 12, 2026 snapshot.
Freshness and provenance
Version
SEAL Instruction Following 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does SEAL Instruction Following measure?
Scale’s SEAL board for precise instruction following on 1,054 private prompts with detailed, specific commands. BenchLM shows it as display-only reference data, excluded from rankings. Scale marked this board deprecated as of January 2025 and no longer adds models.
Which model leads the published SEAL Instruction Following snapshot?
o1 (December 2024) currently leads the published SEAL Instruction Following snapshot with 92.0% scale score. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on SEAL Instruction Following?
The November 12, 2026 snapshot snapshot contains 19 AI models.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.