Tutorials, deep dives and product notes — built for developers.
Cognition’s FrontierCode 2.0 is a new 132-task mergeability benchmark. Interactive leaderboard with best-effort model scores, task pass rates, flag rates, harnesses, and actual evaluation cost per rollout. Updated October 30, 2026.
GPT-6.1 Sol vs GPT-6.1 Terra: Terra costs half as much per token and 1.8× to 2.6× less per Artificial Analysis task at matched effort, while Sol scores 3.7 to 6.5 points higher. Effort-by-effort cost and score comparison, vendor and independent benchmarks, accuracy trade-offs, and a routing guide.
GPT-6.1 Sol replaced GPT-6 Sol after seven days: +4 index points, +12 on Terminal-Bench 4.0, and a cache rate halved to $0.10. It also dropped the "none" reasoning setting, uses 10-30% more output tokens, and is measurably slower.
GPT-6.1 Sol matches Claude Opus 5.5's default index score for 46% less per task and ties it at low effort for 76% less — then stops at index 52, below every Opus effort above default.
DeepSWE v1.1: adds Cognition's SWE-2 Flash (60.1%, provider-reported) alongside Kimi K3.1 Pro (75.5%) and MiniMax M3.1 Flash (73.7%) as provider/model-card runs, separate from Datacurve's mini-swe-agent board. Muse Wonder Full (79.0%) remains the top provider score. Updated Nov 6, 2026.