Models

Costs & Ranking

Compare AI model quality and costs at a glance.

How costs work: The costs shown below represent the approximate cost per turn (reasoning on) once you hit the memory limit (30k tokens). Before the limit, costs are lower.
Quality Ranking
Model
Approx. Cost
1
Deepseek R1
20
2
Claude Sonnet 4
25
3
GPT-4o
30
4
Deepseek v3.1
8
5
Qwen3-Max
10
6
Mistral Large
15
7
Gemini 2.5 Pro
18
8
Kimi K2
5
9
GLM 4.6
6
10
Llama 4 Scout
12