LMCA
A capability assessment with a published score ceiling.
As of 19 Sept 2026, Claude Opus 5 leads LMCA on BenchLeader with 63.3%, ahead of Claude Fable 5 at 60.3%, across 119 model configurations with a published result.
- Published by
- Epoch AI Benchmarking Hub
- Category
- Reasoning
- Index weight
- Reference only
- Models
- 119
- Data as of
- 19 Sept 2026
CC BY 4.0 — Epoch AI, ‘AI Benchmarking Hub’, epoch.ai/benchmarks. Mirrored boards credit their original publishers.
What the test looks like
A broad assessment scored 0 to 100 with a ceiling of 85 for the current version.
How it is scored
Score, as published.
What to keep in mind
The ceiling means top scores compress; new.
- 1Claude Opus 5 (max)63.3%
- 2Claude Fable 5 (max)60.3%
- 3GPT-5.6 Sol59.2%
- 4GPT-5.6 Sol (max)58.4%
- 5Claude Opus 4.8 (max)57.5%
- 6Claude Opus 4.6 (max)55.8%
- 7GPT-5.5 (xhigh)54.3%
- 8GPT-5.5 Pro (xhigh)53.9%
- 9Gemini 3.1 Pro (high)53.8%
- 10Kimi K3 (max)52.7%
- 11GPT-5.6 Terra (max)52.2%
- 12Claude Opus 4.7 (max)52.2%
- 13GPT-5.4 (xhigh)52.0%
- 14Gemini 3.7 Flash (high)50.4%
- 15Muse Spark 1.1 (high)49.9%
119 of 119
| # | ||||
|---|---|---|---|---|
| 1 | 63.3% | 69.9 | 2026-07-24 | |
| 2 | 60.3% | 65.8 | 2026-06-09 | |
| 3 | 59.2% | – | 2026-07-09 | |
| 4 | 58.4% | 68.8 | 2026-07-09 | |
| 5 | 57.5% | 64.2 | 2026-05-28 | |
| 6 | 55.8% | 59.2 | 2026-02-05 | |
| 7 | 54.3% | 67.6 | 2026-04-23 | |
| 8 | 53.9% | 67.1 | 2026-04-23 | |
| 9 | 53.8% | 60.7 | 2026-02-19 | |
| 10 | 52.7% | 67.2 | 2026-07-16 | |
| 11 | 52.2% | 65.0 | 2026-07-09 | |
| 12 | 52.2% | 64.3 | 2026-04-16 | |
| 13 | 52.0% | 65.4 | 2026-03-05 | |
| 14 | 50.4% | 64.6 | 2026-08-13 | |
| 15 | 49.9% | – | 2026-07-09 | |
| 16 | 49.3% | 60.5 | 2026-06-30 | |
| 17 | 48.5% | 60.2 | 2026-07-09 | |
| 18 | 48.5% | 64.6 | 2026-08-12 | |
| 19 | 48.4% | 64.1 | 2026-08-05 | |
| 20 | 47.1% | 63.6 | 2026-05-19 | |
| 21 | 46.5% | 57.6 | 2026-02-17 | |
| 22 | 46.2% | 57.7 | 2026-08-02 | |
| 23 | 45.8% | 63.9 | 2026-06-16 | |
| 24 | 45.2% | 61.0 | 2026-07-08 | |
| 25 | 44.9% | 61.7 | 2026-07-21 | |
| 26 | 44.5% | 57.7 | 2025-11-24 | |
| 27 | 44.0% | 61.0 | 2026-05-19 | |
| 28 | 43.9% | 62.0 | 2025-12-11 | |
| 29 | 43.9% | 58.7 | 2025-11-13 | |
| 30 | 43.1% | 58.7 | 2025-12-17 | |
| 31 | 42.5% | 63.3 | 2026-04-20 | |
| 32 | 41.7% | 54.4 | 2026-07-31 | |
| 33 | 41.2% | 64.2 | 2026-04-24 | |
| 34 | 40.8% | 55.7 | 2026-03-17 | |
| 35 | 40.0% | 58.6 | 2025-08-07 | |
| 36 | 39.7% | 53.0 | 2025-04-16 | |
| 37 | 39.3% | 52.7 | 2026-04-02 | |
| 38 | 38.8% | 54.3 | 2025-09-29 | |
| 39 | 38.7% | – | 2026-02-17 | |
| 40 | 38.5% | 54.5 | 2025-06-10 | |
| 41 | 38.3% | 58.2 | 2026-04-17 | |
| 42 | 37.9% | 56.9 | 2026-02-13 | |
| 43 | 37.6% | 44.3 | 2026-07-21 | |
| 44 | 37.6% | 59.9 | 2026-06-02 | |
| 45 | 37.4% | 52.9 | 2025-05-22 | |
| 46 | 37.3% | 60.5 | 2026-04-20 | |
| 47 | 37.1% | 51.9 | 2025-08-05 | |
| 48 | 36.9% | 48.3 | 2026-06-04 | |
| 49 | 36.9% | 54.6 | 2026-03-17 | |
| 50 | 36.4% | – | 2026-02-16 | |
| 51 | 35.9% | 62.1 | 2026-04-24 | |
| 52 | 35.0% | 49.1 | 2026-03-03 | |
| 53 | 34.8% | 54.5 | 2025-06-17 | |
| 54 | 34.5% | 44.7 | 2026-04-22 | |
| 55 | 34.2% | 55.3 | 2025-08-07 | |
| 56 | 34.0% | 53.0 | 2026-02-24 | |
| 57 | 33.7% | 57.5 | 2026-06-01 | |
| 58 | 33.1% | 58.5 | 2026-03-31 | |
| 59 | 32.2% | 48.1 | 2026-02-24 | |
| 60 | 31.0% | 44.5 | 2026-04-27 |
Cite as: BenchLeader, “LMCA leaderboard”, https://www.benchleader.com/benchmarks/lmca, data as of 19 Sept 2026.
LMCA: questions
- What does LMCA measure?
- A broad assessment scored 0 to 100 with a ceiling of 85 for the current version. Scores are reported in percent of tasks solved; higher is better.
- Which AI model leads LMCA?
- Claude Opus 5 leads LMCA with 63.3% as of 19 Sept 2026, ahead of Claude Fable 5 at 60.3%.
- How many models have LMCA results?
- 119 model configurations have a LMCA result on BenchLeader, all taken from Epoch AI Benchmarking Hub.
- Who runs LMCA and how often is it updated?
- LMCA is published by Epoch AI Benchmarking Hub. BenchLeader re-reads the published results every morning and records the date each result was published.
- Does LMCA count toward the BenchLeader Index?
- No. LMCA is shown for reference but left out of the composite index.