Cybench
Integrity rank #43 of 61 · 19 models scored · top score 93 · Claude Opus 4.6
unknown
Strongest on discrimination, weakest on contamination resistance. Caveat: scores come from an inconsistent mix of harnesses, so head-to-head comparisons are not apples-to-apples.
Reported scores — protocols may differ. How we compare
| # | Model | Score | Evidence | Measured |
|---|---|---|---|---|
| 1 | Claude Opus 4.6 | 93 | self-reported· optimizedT1 | 2026-02-05 |
| 2 | Claude Opus 4.5 | 82 | self-reported· optimizedT1 | 2025-11-24 |
| 3 | Claude Sonnet 4.5 | 60 | self-reported· optimizedT1 | 2025-09-29 |
| 4 | Grok 4 | 43 | self-reported· optimizedT1 | 2025-07-09 |
| 5 | Claude Opus 4.1 | 42 | unverified· optimizedT1 | 2025-08-05 |
| 6 | Claude Opus 4 | 38 | unverified· optimizedT1 | 2025-05-22 |
| 7 | Claude Sonnet 4 | 35 | unverified· optimizedT1 | 2025-05-22 |
| 8 | Grok 4 Fast | 30 | self-reported· optimizedT1 | 2025-09-19 |
| 9 | o3-mini | 22.5 | unverified· optimizedT1 | 2025-01-31 |
| 10 | Claude 3.7 Sonnet | 20 | unverified· optimizedT1 | 2025-02-24 |
| 11 | Claude 3.5 Sonnet | 17.5 | unverified· optimizedT1 | 2024-06-20 |
| 12 | GPT-4.5 | 17.5 | unverified· optimizedT1 | 2025-02-27 |
| 13 | GPT-4o (Nov 2024) | 12.5 | unverified· optimizedT1 | 2024-05-13 |
| 14 | Claude 3 Opus | 10 | unverified· optimizedT1 | 2024-03-04 |
| 15 | o1-mini | 10 | unverified· optimizedT1 | 2024-09-12 |
| 16 | o1-preview | 10 | unverified· optimizedT1 | 2024-09-12 |
| 17 | Llama 3.1-405B | 7.5 | unverified· optimizedT1 | 2024-07-23 |
| 18 | Mixtral 8x22B | 7.5 | unverified· optimizedT1 | 2024-04-17 |
| 19 | Llama 3-70B | 5 | unverified· optimizedT1 | 2024-04-18 |
Continue from this leaderboard into the tightest head-to-head reads.
Weighted composite. Each dimension opens to the source-backed sub-signals it is computed from.
Claims drawn from cited facts, not live model generation.
This benchmark scores 19 models, with a best‑to‑worst score spread of 88.0, indicating a wide separation that supports trustworthy rank differences.
2 cited facts
B-grade at 79 and ranked 35 of 51 — most of the tracked set scores cleaner. Usable for direction; a narrow margin measured here should not settle a model comparison on its own. Contamination narrows how much to trust top scores.
5 cited facts
This benchmark's harness comparability is mixed, meaning scores are not directly comparable and should not be used for head-to-head comparisons across different harnesses. The test set privacy is unknown, so the risk of score inflation due to data exposure cannot be ruled out, which calls for extra scrutiny of any high performance claims for this benchmark. This benchmark has an unknown contamination history and a high contamination window ratio, further increasing the need for caution when interpreting its scores.
4 cited facts
Absent a stated purpose, there is no saying what a score here licenses. It can order entrants against each other but cannot support any capability claim. No construction record exists, so difficulty and item provenance go unchecked; within-table comparisons still stand, but any external claim built on this score rests on an unaudited test. What these tasks are meant to stand in for is unrecorded, and without that the distance between test performance and any real-world skill cannot be assessed. A critical caveat is that the absence of documented limitations means it is unclear where this benchmark may mislead or fail to capture important aspects.
4 cited facts
Cybench: Cybench as reported in Epoch AI's Capabilities Index CSV.
Claude Opus 4.6 leads Cybench at 93 — self-reported by the lab. The full leaderboard above lists every recorded measurement, not just the headline number.
19 models have recorded scores on Cybench, spanning a score spread of 88.
tensor.news grades Cybench B for integrity (score 79/100), ranking #43 of 61 benchmarks we assess across discrimination, saturation, contamination resistance, harness comparability, and freshness.
No — Cybench still has headroom and continues to discriminate between models rather than bunching them at the ceiling.
Its test set is unknown, so contamination risk is on the table: a high score may partly reflect training-data overlap rather than capability.
Only partly — scores come from a mix of harnesses and protocols, so a head-to-head on Cybench is not fully apples-to-apples.
Every Cybench measurement is source-backed and tagged reproduced, self-reported, or unverified; the underlying record cites epoch.ai/data/eci_benchmarks.csv.