LLMs performance reference - Atoti Hub

The following LLMs have been tested against three criteria:

Tool usage: Does the model use tools correctly and consistently?
Answer depth: Can the model handle medium to complex tasks beyond basic UI actions?
Truthfulness: Does the model provide accurate answers without hallucinating?

Provider	Model name	Version	Tool Usage	Answer Depth	Truthfulness
Anthropic	Opus	4.5 and up	🟢	🟢	🟢
Anthropic	Sonnet	4.5 and up	🟢	🟢	🟢
Anthropic	Haiku	4.5 and up	🟢	🔴	🟢
Mistral	Pixtral	Large	🟠	🟢	🔴
Mistral	Small	3.2	🟢	🔴	🟢
OpenAI	GPT	4o	🟢	🟠	🟢

🟢 Good 🟠 Partial 🔴 Poor

⌘I