FrontierCode
Integrity rank #25 of 65 · 42 models scored · top score 54.64 · Claude Opus 5.5
unknown
Strongest on discrimination, weakest on contamination resistance. Caveat: the public test set does not rule out training-data contamination.
Reported scores — protocols may differ. How we compare
| # | Model | Score | Evidence | Measured |
|---|---|---|---|---|
| 1 | Claude Opus 5.5 | 54.64 | unverifiedT1 | 2026-09-22 |
| 2 | Claude Fable 5 | 53.48 | unverifiedT1 | 2026-06-09 |
| 3 | Claude Opus 5 | 53.38 | unverifiedT1 | 2026-07-24 |
| 4 | GPT-6 Astra | 53.26 | unverifiedT1 | 2026-09-03 |
| 5 | Claude Sonnet 5.5 | 52.09 | unverifiedT1 | 2026-09-28 |
| 6 | Claude Fable 5.1 | 50.91 | unverifiedT1 | 2026-09-01 |
| 7 | GPT-6.1 Sol | 50.23 | unverifiedT1 | 2026-09-29 |
| 8 | GPT-6 Sol | 49.27 | unverifiedT1 | 2026-09-22 |
| 9 | Grok 4.6 | 48.01 | unverifiedT1 | 2026-08-12 |
| 10 | Grok 4.7 | 47.59 | unverifiedT1 | 2026-09-21 |
| 11 | Gemini 4 Argon | 47.55 | unverifiedT1 | 2026-09-30 |
| 12 | GPT-5.6 Sol | 47.49 | unverifiedT1 | 2026-07-09 |
| 13 | Claude Opus 4.8 | 46.5 | unverifiedT1 | 2026-05-28 |
| 14 | Claude Haiku 5.5 | 46.36 | unverifiedT1 | 2026-10-07 |
| 15 | Kimi K3 | 44.17 | unverifiedT1 | 2026-07-16 |
| 16 | DeepSeek V4.1 Flash | 43.87 | unverifiedT1 | 2026-09-09 |
| 17 | Gemini 3.7 Flash | 43.59 | unverifiedT1 | 2026-08-13 |
| 18 | GPT-5.5 | 42.96 | unverifiedT1 | 2026-04-23 |
| 19 | Claude Sonnet 5 | 42.73 | unverifiedT1 | 2026-06-30 |
| 20 | Grok 4.5 | 42.44 | unverifiedT1 | 2026-07-08 |
| 21 | GPT-6 Luna | 42.42 | unverifiedT1 | 2026-09-22 |
| 22 | GPT-5.6 Terra | 41.31 | unverifiedT1 | 2026-07-09 |
| 23 | Gemini 3.8 Flash | 41.19 | unverifiedT1 | 2026-09-02 |
| 24 | GLM-5.3 | 40.14 | unverifiedT1 | 2026-08-14 |
| 25 | GPT-5.6 Luna | 39.81 | unverifiedT1 | 2026-07-09 |
| 26 | Mistral Large 4 | 39.43 | unverifiedT1 | 2026-10-06 |
| 27 | Claude Opus 4.7 | 38.54 | unverifiedT1 | 2026-04-16 |
| 28 | Gemini 3.6 Flash | 34.37 | unverifiedT1 | 2026-07-21 |
| 29 | GLM-5.3-Flash | 31.83 | unverifiedT1 | 2026-08-20 |
| 30 | Kimi K2.7 Code | 30.06 | unverifiedT1 | 2026-06-12 |
| 31 | DeepSeek V4 Pro 0813 | 28.55 | unverifiedT1 | 2026-08-13 |
| 32 | GPT-5.4 Mini | 27.04 | unverifiedT1 | 2026-03-17 |
| 33 | Claude Opus 4.6 | 26.64 | unverifiedT1 | 2026-02-05 |
| 34 | GLM-5.2 | 24.5 | unverifiedT1 | 2026-06-16 |
| 35 | Claude Sonnet 4.6 | 24.31 | unverifiedT1 | 2026-02-17 |
| 36 | DeepSeek V4 Flash 0731 | 18.8 | unverifiedT1 | 2026-07-31 |
| 37 | DeepSeek-V4-Pro | 17.64 | unverifiedT1 | 2026-04-24 |
| 38 | MiniMax-M3 | 14.7 | unverifiedT1 | 2026-06-01 |
| 39 | Inkling | 14 | unverifiedT1 | 2026-07-15 |
| 40 | Nemotron 3 Ultra | 13.61 | unverifiedT1 | 2026-06-04 |
| 41 | Qwen3.7-Plus | 10.2 | unverifiedT1 | 2026-06-02 |
| 42 | Mistral Medium 3.5 | 8 | unverifiedT1 | 2026-04-28 |
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.
The recorded metadata does not specify what this benchmark measures.
1 cited fact
This benchmark has scores from 42 tracked models, with a recorded score range spanning 46.64.
2 cited facts
Its Benchmark Integrity Index score is 87, with grade A and rank 25 out of 65 tracked benchmarks.
4 cited facts
It is not flagged as saturated. The top recorded score is 54.64, with 45.36 points of headroom. The computed top-score cluster contains 4 models.
4 cited facts
The harness classification is consistent harness. Test-set privacy is recorded as unknown. The recorded contamination-window ratio is 0.738.
3 cited facts
FrontierCode: FrontierCode as reported in Epoch AI's Capabilities Index CSV.
Claude Opus 5.5 leads FrontierCode at 54.64. The full leaderboard above lists every recorded measurement, not just the headline number.
42 models have recorded scores on FrontierCode, spanning a score spread of 46.64.
tensor.news grades FrontierCode A for integrity (score 87/100), ranking #25 of 65 benchmarks we assess across discrimination, saturation, contamination resistance, harness comparability, and freshness.
No — FrontierCode 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.
Scores are reported under a consistent harness, so comparisons on FrontierCode are reasonably apples-to-apples.
Every FrontierCode measurement is source-backed and tagged reproduced, self-reported, or unverified; the underlying record cites epoch.ai/data/eci_benchmarks.csv.