reasoning benchmark
Agents' Last Exam Leaderboard
Agents' Last Exam is a benchmark for AI agents on hard, long-horizon tasks, reported with and without tool access.
Updated Sep 7, 2026
Models16
Model coverage16
MetricScore
EvidenceB
Agents' Last Exam Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelOP | Score59.30% | Percentile100.00% | Participants16 | EvidenceC | Evaluated |
| Rank02 | ModelOP | Score52.70% | Percentile93.33% | Participants16 | EvidenceC | Evaluated |
| Rank03 | ModelAC | Score52.40% | Percentile86.67% | Participants16 | EvidenceC | Evaluated |
| Rank04 | ModelAC | Score51.20% | Percentile80.00% | Participants16 | EvidenceC | Evaluated |
| Rank05 | ModelAC | Score51.20% | Percentile73.33% | Participants16 | EvidenceC | Evaluated |
| Rank06 | ModelOP | Score50.40% | Percentile66.67% | Participants16 | EvidenceC | Evaluated |
| Rank07 | ModelOP | Score50.30% | Percentile60.00% | Participants16 | EvidenceC | Evaluated |
| Rank08 | ModelAC | Score42.90% | Percentile53.33% | Participants16 | EvidenceC | Evaluated |
| Rank09 | ModelBY | Score41.40% | Percentile46.67% | Participants16 | EvidenceC | Evaluated |
| Rank10 | ModelZA | Score28.50% | Percentile40.00% | Participants16 | EvidenceC | Evaluated |
| Rank11 | ModelDE | Score27.30% | Percentile33.33% | Participants16 | EvidenceC | Evaluated |
| Rank12 | ModelGO | Score26.30% | Percentile26.67% | Participants16 | EvidenceC | Evaluated |
| Rank13 | ModelZA | Score26.30% | Percentile20.00% | Participants16 | EvidenceC | Evaluated |
| Rank14 | ModelDE | Score25.70% | Percentile13.33% | Participants16 | EvidenceC | Evaluated |
| Rank15 | ModelDE | Score25.20% | Percentile6.67% | Participants16 | EvidenceC | Evaluated |
| Rank16 | ModelTE | Score22.80% | Percentile0.00% | Participants16 | EvidenceC | Evaluated |
Agents' Last Exam Highlights
The leading models and scores on this benchmark.
Agents' Last Exam Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for Agents' Last Exam
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What is Agents' Last Exam?
What Agents' Last Exam measures and how its scores work.
Agents' Last Exam is a reasoning benchmark for AI agents.
It measures sustained reasoning, planning, and tool use on hard, long-horizon tasks using Score, reported as a ratio, with and without tool access.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- Agents' Last Exam
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- overall
LLM Stats. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about Agents' Last Exam.
Which model scores highest on Agents' Last Exam?
GPT-6 Astra is currently ranked first with 59.30%.
What are the top three models on Agents' Last Exam?
The current leaders are GPT-6 Astra (59.30%), GPT-5.6 Sol (52.70%), and Qwen3.8 Max (52.40%).
Which Agents' Last Exam model has the lowest official input price?
GLM-5.3-Flash has the lowest matched official input price at $0.08 input / $0.25 output per 1M tokens.
Which models are fastest among Agents' Last Exam results?
The fastest matched records are GLM-5.3 (472.69 tok/s via FriendliAI), Gemini 3.7 Flash (117.86 tok/s via Google), and GPT-5.6 Terra (99.35 tok/s via OpenAI).
Does the highest score result prove overall model quality?
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.