Artificial Analysis Coding Agent Index (AA Coding Agents)
We show this table for reference; we do not rank on it.
A display-only Artificial Analysis leaderboard for coding-agent systems, combining agent harnesses, host models, and execution settings across software-engineering benchmarks.
Index score on AA Coding Agents — Source captured 2026-09-23; evaluation dates not inferred
We mirror the published index score view for AA Coding Agents. Claude Code - Fable 5.1 (max) (with fallback) leads the public snapshot at 62.2%, followed by Devin Fusion CLI - Claude Fable 5.1 XHigh + SWE-2 Medium (61.7%) and Codex - GPT-6 Astra (max) (61.6%). We do not use these results to rank models overall.
Claude Code - Fable 5.1 (max) (with fallback)
Anthropic
Devin Fusion CLI - Claude Fable 5.1 XHigh + SWE-2 Medium
Devin Fusion CLI
Codex - GPT-6 Astra (max)
OpenAI
19 modelsCodingCurrentDisplay onlyUpdated Source captured 2026-09-23; evaluation dates not inferred
Index score table (19 models)
ScoreHow AA Coding Agents is shown here
BenchLM mirrors the Artificial Analysis Coding Agent Index v1.1 page as a display-only agent leaderboard. The source compares coding-agent variants across DeepSWE v1.1, SWE-Atlas-QnA, Terminal-Bench v4 and reports the average pass@1 index alongside cost, token, and execution-time metadata.
AA Coding Agents is separate from BenchLM model-only rankings. Its rows combine an agent harness, a host model, execution settings, and provider routing, so BenchLM treats the index as external system evidence rather than a weighted base-model benchmark. Component benchmark availability can vary by row in the source payload.
Snapshot
The published AA Coding Agents snapshot places Claude Code - Fable 5.1 (max) (with fallback) first at 62.2%. The third row is 0.6 points behind. The broader top-10 range is 8.6 points, so many of the published results sit in a relatively narrow band.
19 models have been evaluated on AA Coding Agents. The benchmark falls in the Coding category. AA Coding Agents is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About AA Coding Agents
Year
2026
Tasks
Composite over DeepSWE, Terminal-Bench v2, and SWE-Atlas-QnA
Format
Average pass@1 index
Difficulty
Real-world coding-agent workflows
BenchLM mirrors the Artificial Analysis Coding Agent Index v1.1 page as a display-only external leaderboard. The source combines DeepSWE, Terminal-Bench v2, and SWE-Atlas-QnA component scores and publishes cost, token, and execution-time metadata. Rows are coding-agent systems rather than pure base-model results.
Freshness and provenance
Version
AA Coding Agents 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does AA Coding Agents measure?
A display-only Artificial Analysis leaderboard for coding-agent systems, combining agent harnesses, host models, and execution settings across software-engineering benchmarks.
Which model leads the published AA Coding Agents snapshot?
Claude Code - Fable 5.1 (max) (with fallback) currently leads the published AA Coding Agents snapshot with 62.2% index score. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on AA Coding Agents?
The Source captured 2026-09-23; evaluation dates not inferred snapshot contains 19 AI models.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 2,000+ readers.
One email each week. Unsubscribe anytime.