LLM Effort Levels Compared

Last update: October 1st, 2026

GPT-6.1-Sol: Low vs Medium vs High

Tested with Codex CLI.

Effort	Total points (max 80)	Avg cost per prompt	Avg time per prompt	vs previous level
Low	65.36	$0.10	03:30	—
Medium	67.09	$0.14	03:53	+1.73 pts · +46% cost · +00:23 time
High Sweet spot	68.1	$0.22	07:26	+1.01 pts · +59% cost · +03:33 time

Tested with Claude Code.

Effort	Total points (max 80)	Avg cost per prompt	Avg time per prompt	vs previous level
Low	60.57	$0.37	01:13	—
Medium	65.5	$0.56	02:04	+4.93 pts · +53% cost · +00:51 time
High Sweet spot	67.41	$0.79	03:10	+1.91 pts · +42% cost · +01:06 time

Tested with Codex CLI.

Effort	Total points (max 80)	Avg cost per prompt	Avg time per prompt	vs previous level	Video
Medium Sweet spot	66.03	$0.69	03:08	—
High	66.35	$0.96	04:34	+0.32 pts · +39% cost · +01:26 time

Tested with Claude Code.

Effort	Total points (max 80)	Avg cost per prompt	Avg time per prompt	vs previous level	Video
Medium	60.54	$0.20	00:58	—
High Sweet spot	64.31	$0.34	01:53	+3.77 pts · +67% cost · +00:55 time

Tested with Codex CLI.

Effort	Total points (max 80)	Avg cost per prompt	Avg time per prompt	vs previous level
Medium	43.74	$0.01	02:13	—
High	53.04	$0.01	03:51	+9.30 pts · +15% cost · +01:38 time
Xhigh	56.4	$0.03	10:08	+3.36 pts · +122% cost · +06:17 time
Max Sweet spot	59.07	$0.03	11:03	+2.67 pts · +12% cost · +00:55 time

Tested with OpenCode.

Effort	Total points (max 80)	Avg cost per prompt	Avg time per prompt	vs previous level	Video
High Sweet spot	54.35	$0.03	03:17	—
Max	53.31	$0.05	05:15	-1.04 pts · +61% cost · +01:58 time

The sweet spot is the lowest effort level that scores within 1 point of the model's best result.
Score bars start at 30 points, so the differences between effort levels are easier to see.
Each score measures a model-and-harness configuration, not the model in isolation.
Prices are calculated with API costs. Models tested on a subscription plan show N/A.
Full methodology and scoring formulas are explained in this article.

See every tested model side by side on the full LLM Coding Leaderboard.