Compare AI models

Specifications
Parameters
—0.5B/1.5B/3B/7B/14B/32B
Context window
128k128k
API pricing
Input price
$0.15—
Output price
$0.60—
Cached input price
$0.075—
Cheapest input
$0.15Azure$0.66Cloudflare
Cheapest output
$0.60Azure$1.00Cloudflare
$60.00OpenAI list price$182Cloudflare
Benchmarks
HumanEval
87.2%92.7%
MATH
70.2%76.4%
MMLU
82%77.6%
Overview
CompanyOpenAIQwen
Release dateJul 18 2024Sep 19 2024
AccessClosedOpen Weight
Model detailsView modelView model

Frequently asked questions

Qwen2.5-Coder leads GPT-4o mini on 2 of the 3 benchmarks they both report (HumanEval, MATH, MMLU). Only GPT-4o mini has a verified first-party price for an available API: $0.15 per million input tokens and $0.60 per million output tokens. No currently available pay-as-you-go API rate is tracked for Qwen2.5-Coder. GPT-4o mini shipped 63 days before Qwen2.5-Coder, so benchmark comparisons should account for the intervening progress.

Context windows are 128k (GPT-4o mini) vs 128k (Qwen2.5-Coder). GPT-4o mini is closed, while Qwen2.5-Coder is open weight.

On HumanEval, Qwen2.5-Coder leads at 92.7% vs GPT-4o mini at 87.2%. On MATH, Qwen2.5-Coder leads at 76.4% vs GPT-4o mini at 70.2%. On MMLU, GPT-4o mini leads at 82% vs Qwen2.5-Coder at 77.6%.