LLM Reference
activeCodingAgents

FrontierCode 1.1 Main

Metric: Mean@5 score (higher is better)Introduced: 2026

Cognition's FrontierCode 1.1 Main evaluation measures whether coding agents produce mergeable changes on the 100 hardest tasks from its 150-task Extended set. It reports a mean@5 composite of held-out test performance and weighted code quality; Main remains distinct from the full Extended variant.

Models ranked

2

tracked on this benchmark

Score band

53.4 – 43.6

best → lowest tracked

Snapshot trend

-9.80

Jul 24 → Aug 13 · 1 models

Leaderboard

Tracked models ranked by Mean@5 score (higher is better).

Compare candidates
#Model variant and provenanceScore
1
Claude Opus 5

Configuration: Medium effort

Version: FrontierCode 1.1 MainHarness: 100-task mean@5 composite of held-out-test performance and weighted code qualityEvaluator: CognitionObserved: Jul 24, 2026Confidence: confirmedSource

Notes: Source: p.150, Figure 8.3. Medium is the reported Opus 5 setting; preserve it and keep Main separate from Extended.

53.4
2
Gemini 3.7 Flash
Version: FrontierCode 1.1 MainHarness: Not recordedEvaluator: Not recordedObserved: Aug 13, 2026Confidence: Not recordedSource
43.6

How to read this benchmark

This benchmark scores models where higher is better. Use scores for directional filtering and shortlisting, not universal quality ranking; then validate pricing, context window, provider availability, and fit for your workload.

Trust this score when

  • There is a fresh timestamped snapshot (or multiple snapshots) for this benchmark.
  • The model list covers the same version family you can actually deploy today.
  • Top candidates overlap with your required routing and feature requirements.

Be cautious when

  • There is only one benchmark snapshot or the dataset appears stale.
  • The benchmark metric direction is opposite of your decision objective.
  • The score difference between options is narrow and likely within implementation variance.

Related benchmarks

Last reviewed: Jul 26, 2026

Resources