OSWorld 2.0 vs Effective Cost
X = effective cost (log). Y = OSWorld 2.0. Official OSWorld 2.0 leaderboard only: 108 long-horizon desktop tasks, 500-step budget, full task set, binary completion accuracy, best reasoning/tool configuration per model. Partial-credit scores and provider-reported launch figures are excluded.
OSWorld 2.0 vs Effective Cost
OSWorld 2.0 vs Effective Cost
X = effective cost (log). Y = OSWorld 2.0. Official OSWorld 2.0 leaderboard only: 108 long-horizon desktop tasks, 500-step budget, full task set, binary completion accuracy, best reasoning/tool configuration per model. Partial-credit scores and provider-reported launch figures are excluded.
Controls:
How to read this chart
Each point is a public model. The chart compares OSWorld 2.0 (%) against Effective Cost (per 1M I/O Tokens), with color showing the model provider.
More / ExperimentalSWE-Bench Verified Benchmark ScoresHistorical data. Vals archived cohort; retired from default Software Engineering domain. Each model's SWE-Bench Verified score. Color = provider.Data: Vals.ai, LLM StatsOpen chartMore / ExperimentalTerminal-Bench 2.0 Benchmark ScoresHistorical data. Superseded by 2.1 and 4.0. Legacy Terminal-Bench 2.0 scores retained for historical comparison. Color = provider.Data: Terminal-BenchOpen chartMore / ExperimentalTerminal-Bench 2.1 Benchmark ScoresHistorical data. Superseded in Programmatic Reasoning by Terminal-Bench 4.0. Each model's Terminal-Bench 2.1 pass@1 score, using Artificial Analysis as canonical and Vals.ai as fallback. Color = provider.Data: Artificial Analysis Terminal-Bench v2.1, Vals.ai Terminal-Bench 2.1, Artificial Analysis Intelligence Index v4.3.1Open chartMore / ExperimentalTerminal-Bench Hard Benchmark ScoresHistorical data. Historical hard subset; not a Terminal-Bench 4.0 result. Each model's Terminal-Bench Hard score. Color = provider.Data: Artificial AnalysisOpen chartMore / ExperimentalARC-AGI-3 Cost EfficiencyX = ARC Prize reported Cost (V3) (log). Y = ARC-AGI-3 %. Each line connects one model's published reasoning-effort levels, so the score-vs-cost tradeoff is visible per model. Defaults to the current model generation. Color = provider.Data: ARC PrizeOpen chartMore / ExperimentalARC-AGI-2 Cost EfficiencyX = ARC Prize reported cost/task (log). Y = ARC-AGI-2 %. Each line connects one model's published reasoning-effort levels, so the score-vs-cost tradeoff is visible per model. Defaults to the current model generation. Color = provider.Data: ARC PrizeOpen chart