See what coding
performance costs.
An interactive view of three coding benchmarks, updated to show GPT-5.6 Terra and Luna at current Standard and Fast API rates.
Explore the benchmarksLuna Standard versus the original launch-chart cost.
Cost on the x-axis. Capability on the y-axis.
Higher and farther left is better. Select a benchmark, inspect individual effort levels, or hide a series to simplify the comparison.
Artificial Analysis Coding Index
General coding-agent capability across a broad task mix.
Terra: Standard 0.8×, Fast 1.6× original cost. Luna: Standard 0.2×, Fast 0.4×. All other costs and every benchmark score are unchanged.
The scores stayed still. Only the price moved.
The original chart values were extracted from the rendered GPT-5.6 launch charts. Terra and Luna costs were then scaled by the exact ratio between their original and current token prices. Because every billing component changed by the same ratio within each mode, the adjustment does not depend on an assumed token mix.
- 01Preserve every original benchmark score and competitor point.
- 02Recalculate Terra and Luna cost coordinates for Standard and Fast.
- 03Validate source values against chart geometry and benchmark tables.
Scope note: short-context pricing only. Fast mode does not support long context. Terminal-Bench Terra and Luna chart maxima differ from the launch-page table; this explorer preserves the chart-embedded values.