The Best AI Models for Math

Compare the top frontier models for **Math** tasks, balanced for intelligence and API cost.

Top Frontier Leaders

Frontier Model	LMArena Elo	API Cost (1M)	Throughput
qwen3.5-27b	1430	$1.95e-7	59
minimax-m2.5	1422	$1.2e-7	21
qwen3-235b-a22b-instruct-2507	1420	$7.1e-8	2
qwen3.5-flash	1411	$6.5e-8	86
gpt-oss-120b	1384	$3.9e-8	52
gpt-oss-20b	1337	$3e-8	3
gemma-3n-e4b-it	1259	$2e-8	13

*These models represent the Pareto Frontier (optimal cost-to-performance).*