← All guidesAt a glance
Launched alongside Grok 3 in February 2025 as the efficient tier of the generation - less world knowledge than its big sibling, but with the same reasoning training applied to a smaller, faster model.
On the All-inclusive plan a Grok 3 mini reply costs 1 credit.
Benchmarks
- AIME 2024 (competition maths): 95.8% in reasoning mode - above many far larger models.
- LiveCodeBench (code generation): 80.4% - edging out full Grok 3 on this one.
Best at on Magnus
Quick asks, maths-flavoured questions, and cheap chatty agents that keep the Grok personality without the Grok price. Pair it with the humour slider for a lively 1-credit sidekick.