Best of
The models that get the number right.
Math picks straight from the live math lane (AIME and MATH-500 among the named benchmarks), with prices beside each pick.
Last reviewed July 16, 2026
Math is the most verifiable thing a model does: the answer is right or it is not, and the public benchmarks measure exactly that. The picks below follow the math lane of the live rankings, from the frontier reasoners that top it to the budget models close enough to run on every problem. Reasoning models bill their thinking tokens, so the live price beside each pick is part of the decision.
The picks
$30 in / $180 out per M tokens · 1.1M context
The current leader on the competition-math benchmarks; the pick when a single hard problem has to come back right.
Best for: The hardest problems, worth the price.
$10 in / $50 out per M tokens · 1M context
Within a rounding error of the lane's top, with derivations legible enough to check step by step instead of taking on faith.
Best for: Proofs and work you need to verify.
$2 in / $6 out per M tokens · 500K context
xAI's reasoner sits near the top of the math lane and holds up on olympiad-style problems where the trick matters more than the grind.
Best for: Competition-style problem solving.
$0.43 in / $0.87 out per M tokens · 1.0M context
Close enough to the frontier on the math lane to handle most coursework and modeling, at prices that let you run it on everything.
Best for: Volume: homework checks, practice, iteration.
$0.78 in / $3.90 out per M tokens · 262K context
Strong MATH-500 standing with visible reasoning traces and open weights; a second opinion cheap enough to ask always.
Best for: Step-by-step work at open-weight prices.
Frequently asked
Judge them on your own prompt
Every pick on this page runs in one idapt chat. Start free with a daily allowance, no credit card.
Try idapt free