Kiro's GPT-5.6 Luna credit multiplier is about 83% off. I checked how Terra / Luna appear after the change
This page has been translated by machine translation. View original
Introduction
On August 1, 2026, the credit multipliers for GPT-5.6 Terra / Luna shown on Kiro's model selection screen were updated. The displayed values before and after the change are as follows. The reduction rate is calculated from the displayed multipliers.
| Model | Before | After | Reduction |
|---|---|---|---|
gpt-5.6-terra |
1.20x credits | 1.00x credits | About 17% |
gpt-5.6-luna |
0.60x credits | 0.10x credits | About 83% |
GPT-5.6 Terra / Luna became available in Kiro on July 14, 2026, so this multiplier update came about half a month later. I published an article the day before about the pricing revision for GPT-5.6 Terra / Luna on Amazon Bedrock. The price reduction on the Bedrock side concerns the USD per-token price, which is a different metric from the credit multipliers shown on Kiro's model selection screen. This article covers Kiro's credit multipliers.
What I Verified
Confirming the multiplier change on the CLI model selection screen
I checked the model list shown by /model in Kiro CLI before and after the change. Before the change, GPT-5.6 Terra / Luna were displayed as follows.
gpt-5.6-terra 1.20x credits
gpt-5.6-luna 0.60x credits
Terra went from 1.20x to 1.00x, a reduction of about 17% in the credit multiplier. Luna went from 0.60x to 0.10x, a reduction of about 83%. Both values are calculated from the displayed multipliers.
Among the models displayed after the change, those with a credit multiplier of 1.00x or lower are as follows.
| Model | Credit multiplier |
|---|---|
| auto | 1.00x |
gpt-5.6-terra |
1.00x (1.20x before the change) |
| glm-5 | 0.50x |
| claude-haiku-4.5 | 0.40x |
| deepseek-3.2 | 0.25x |
| minimax-m2.5 | 0.25x |
| minimax-m2.1 | 0.15x |
gpt-5.6-luna |
0.10x (0.60x before the change) |
| qwen3-coder-next | 0.05x |
Confirming the update on the IDE model selection screen as well
Not only in the CLI, but also on the Kiro IDE model selection screen, I was able to confirm the updated multipliers for GPT-5.6 Terra / Luna.
Summary
With its credit multiplier now at 0.10x, GPT-5.6 Luna is a strong candidate when working in Kiro on tasks that do not require high performance. Using a model with a high multiplier for simple tasks tends to increase credit consumption and time required, so you can adopt an approach of trying Luna first and, if you do not get the expected output, switching to a higher-tier model such as Terra, Sol, Sonnet, or Opus. As an entry point for choosing models according to the task rather than fixing one model, Luna now has a multiplier that makes it easy to try.
Reference Links
