Three levels: Provider → Base Model → Reasoning Level. Each base model can have multiple reasoning levels. Higher levels are usually more accurate but slower and costlier. Click any card to drill into fixture-by-fixture results.

Anthropic

A

Arcee-ai

Deepseek

Google

I

Ibm-granite

I

Inception

I

Inclusionai

ling-2.6-flash

Text Output
JSON Schema
L

Liquid

lfm-2-24b-a2b

Text Output
JSON Schema

Meta

Minimax

M

Mistralai

M

Moonshotai

N

Nvidia

Openai

P

Poolside

P

Prism-ml

Qwen

S

Sakana

S

Stealth

T

Tencent

T

Thinkingmachines

X

X-ai

X

Xiaomi

Z

Z-ai