The Best AI Models for Instruction following

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

Top Frontier Leaders

Frontier Model	LMArena Elo	API Cost (1M)	Throughput
deepseek-v3.1-thinking	1418	$1.5e-7	14
qwen3-235b-a22b-instruct-2507	1416	$7.1e-8	2
qwen3.5-flash	1380	$6.5e-8	86
glm-4.7-flash	1351	$6e-8	74
gpt-oss-120b	1326	$3.9e-8	52
mistral-small-3.1-24b-instruct-2503	1294	$3e-8	1
gemma-3n-e4b-it	1280	$2e-8	13

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