Frontier snapshots

Generated 2026-09-01T00:30:00Z · raw data

Models across all frontiers

★ = on frontier · ≈ = within 10% of frontier · hover a row to highlight that model in every chart; hover a chart point to highlight its row.

modelAAbest $/turnsub·costsub·mathsub·speedmetered·costmetered·mathmetered·speedsub·code-costsub·code-mathmetered·code-mathsub·code-speed
gpt-5.6-sol61$0.1740
gpt-5.6-luna52$0.0070··
glm-5.360$0.0664····
deepseek-v4-flash-073152$0.0095······
gpt-oss-120b24$0.0006·······
opus-563$0.1711········
claude-fable-562$0.3421········
gpt-5.6-terra57$0.0696·······
gpt-oss-20b15$0.0005········
muse-spark-1.2-contributor57$0.0019·········
qwen3.8-max52$0.0744·········
qwen3-30b-a3b-25079$0.0008·········
qwen3.8-2.4t-a95b52$0.0744·········
qwen3-coder-30b14$0.0012··········
glm-5.3-flash57$0.0037··········
longcat-2.034$0.0061··········
muse-spark-1.257$0.0457··········

Charts

Interactive: hover for exact values, zoom/pan by drag. † = assumed cap.

ClinePass cap assumed $35; $20-tier sub caps modeled 8× price — †-flagged. Sources: AA Intelligence Index v4.1.1 (current scale), vals.ai SWE-bench Verified (independent same-harness), AA-run GPQA Diamond, OpenRouter catalog, published subscription tables.