Context-Bench Skills: leaderboard
Letta agentic context engineering benchmark: skills suite testing task completion and skill use capabilities across multi-step agentic workflows.
Metric: Task Completion (%). Source: leaderboard.letta.com. Saturation forecast: Around January 2027. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 (2025-12-11) (xHigh) | 85.31 | |
| 2 | GPT-5.2 (2025-12-11) (High) | 84.47 | |
| 3 | Claude Opus 4.6 | 81.32 | #63 |
| 4 | Claude Sonnet 4.6 | 79.08 | #90 |
| 5 | GPT-5.2 (2025-12-11) (Medium) | 77.55 | |
| 6 | Claude Sonnet 4.5 | 76.5 | #142 |
| 7 | Claude Opus 4.5 (20251101) | 75.54 | #74 |
| 8 | DeepSeek V3 Chat | 75.33 | #447 |
| 9 | Claude Opus 4.1 (20250805) | 74.4 | #132 |
| 10 | Gemini 3 Pro | 72.73 | #80 |
| 11 | DeepSeek Reasoner | 70.48 | #242 |
| 12 | GPT-5 | 70.2 | #98 |
| 13 | Claude Haiku 4.5 (20251001) | 69.7 | #261 |
| 14 | GPT-5 Mini (2025-08-07) | 68.8 | #172 |
| 15 | GLM-4.6 | 65.9 | #252 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=context-bench-skills · How It Works · Data refreshed daily, snapshot 2026-10-20.