LLM Stats (Agents' Last Exam): leaderboard
LLM Stats mirror of Agents' Last Exam, a long-horizon agent benchmark on hard tasks requiring sustained reasoning, planning and tool use; upstream marks the board unverified — every row is a vendor self-report, and results may be reported with or without tool access.
Metric: Score (%). Source: llm-stats.com. Saturation forecast: Around September 2028. 26 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6.1 Terra | 60 | #15 |
| 2 | GPT-6 Astra | 59.3 | #6 |
| 3 | GPT-6 Sol | 56.4 | #19 |
| 4 | GPT-5.6 Sol | 52.7 | #22 |
| 5 | Qwen 3.8 Max | 52.4 | #34 |
| 6 | Qwen3.8-Flash-Next | 51.2 | #76 |
| 7 | GPT-6 Luna | 50.9 | #61 |
| 8 | GPT-5.6 Terra | 50.4 | #40 |
| 9 | GPT-5.6 Luna | 50.3 | #79 |
| 10 | Kimi K3.1 Pro | 43.2 | #17 |
| 11 | Qwen 3.8 27B | 42.9 | #123 |
| 12 | Seed 2.1 Pro | 41.4 | #73 |
| 13 | Gemini 4 Argon | 39.5 | #9 |
| 14 | DeepSeek V4.1 Flash (1017) | 35.4 | #27 |
| 15 | DeepSeek V4.1 Flash | 31.8 | #37 |
No result here: #3 Claude Opus 5.5, #7 GPT-6.1 Sol, #10 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-agents-last-exam · How It Works · Data refreshed daily, snapshot 2026-11-05.