FrontierSWE v2
Proximal's FrontierSWE v2 is a 34-task ultra-long-horizon engineering and research benchmark (system implementation, performance optimization, scientific computing, AI research, and visual reasoning) with a 20-hour budget per trial. The standard protocol runs each model at max reasoning effort in Proximal's agent harness and reports the mean across five trials per task.
Models ranked
1
tracked on this benchmark
Score band
69.9 – 69.9
best → lowest tracked
Snapshot trend
—
need ≥2 snapshots
Leaderboard
Tracked models ranked by Mean score (higher is better).
Configuration: claude-fable-5-2 (max reasoning, Proximal harness, mean of 5 trials)
Notes: Anthropic self-reported (lab-published). System card + launch table, retrieved 2026-11-09. Does not move /best rankings (self-reported / raon rule). System card §8.7 and launch table: 69.9%. Proximal FrontierSWE v2, 34 ultra-long-horizon tasks, 20-hour budget, max reasoning effort in Proximal's own agent harness, mean of five trials per task. Card says Fable 5.2 was first among the models Proximal evaluated. Owner site https://www.frontierswe.com/blog/v2 was captured today but did not list a Fable 5.2 score in the static HTML, so this row stays lab-published (evaluator Anthropic citing Proximal's eval) rather than a scraped owner-board restamp.
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: Nov 9, 2026