LLM ReferenceLLM Reference

Best Free LLMs (2026)

Free-to-use large language models with open weights you can run locally or via free-tier APIs. Ranked by MMLU to surface the most capable free options first.

#ModelInput $/1MOutput $/1M
1Llama 3.1 405B
2DeepSeek V3
Tools
$0.10$0.28
3Qwen2.5 72B
$0.20$0.60
4Qwen2.5 32B Instruct
$0.15$0.15
5Llama 3.1 70B Instruct
$0.40$0.40
6Mixtral 8x22B v0.1
GPU·hr$1.00
7Falcon 180B
8Qwen2.5 14B Instruct
$0.08$0.08
9Qwen2 72B
$0.45$0.65
10Mistral Large 2
VisionTools
$0.48$1.50
11Gemma 2 27B Instruct
GPU·hr$1.00
12Llama 3 70B Instruct
$0.40$0.40
13Mistral NeMo Instruct (2407)
GPU·hr$1.00
14Qwen2.5 7B Instruct
$0.04$0.04
15Qwen2 7B
$0.05$0.15
16Mixtral 8x7B
$0.15$0.20
17Qwen1.5-72B
$0.65$0.90
18DeepSeek Coder V2 Lite
$0.50$0.50
19Mistral 7B Instruct v0.3
Tools
$0.20$0.20
20Qwen1.5-110B
$1.50$1.80