Training & Fine-Tuning Cost Leaderboard

Find the cheapest way to train or fine-tune any AI model.

311 models · 14 methods · 60 GPUs

Inference Leaderboard
🥇#1QLoRA

AlibabaQwen 2.5 7B

Qwen 2.5 · 7.6B

GPU
T4
Time
6.7h
Cost
$0.80
Quality
88%
Fine-tune this model
🥈#2QLoRA

AlibabaQwen 3 8B

Qwen 3 · 8.2B

GPU
T4
Time
7.2h
Cost
$0.86
Quality
88%
Fine-tune this model
🥉#3GaLore

AlibabaQwen 2.5 1.5B

Qwen 2.5 · 1.5B

GPU
T4
Time
1.3h
Cost
$0.16
Quality
97%
Fine-tune this model
#4GaLore

AlibabaQwen 2.5 3B

Qwen 2.5 · 3.1B

GPU
T4
Time
2.7h
Cost
$0.32
Quality
97%
Fine-tune this model
#5QLoRA

MetaLlama 3.1 8B

Llama 3.1 · 8B

GPU
T4
Time
7.0h
Cost
$0.84
Quality
88%
Fine-tune this model
#6ReFT

AlibabaQwen 3 4B

Qwen 3 · 4B

GPU
T4
Time
3.5h
Cost
$0.42
Quality
83%
Fine-tune this model
#7ReFT

MetaLlama 3.2 3B

Llama 3.2 · 3.2B

GPU
T4
Time
2.8h
Cost
$0.34
Quality
83%
Fine-tune this model
#8ReFT

AlibabaQwen 3 32B

Qwen 3 · 32.8B

GPU
H100 SXM
Time
1.9h
Cost
$3.6
Quality
83%
Fine-tune this model
#9GaLore

MetaLlama 3.2 1B

Llama 3.2 · 1.2B

GPU
T4
Time
1.1h
Cost
$0.13
Quality
97%
Fine-tune this model
#10QLoRA

MetaLlama 3 8B

Llama 3 · 8B

GPU
T4
Time
7.0h
Cost
$0.84
Quality
88%
Fine-tune this model
#11QLoRA

MetaLlama 3.1 70B

Llama 3.1 · 70.6B

GPU
H100 SXM
Time
4.1h
Cost
$7.7
Quality
88%
Fine-tune this model
#12QLoRA

MetaHelpSteer2 Llama 3.1 70B

Llama 3.1 · 70.6B

GPU
H100 SXM
Time
4.1h
Cost
$7.7
Quality
88%
Fine-tune this model
#13QLoRA

MetaLlama 3.1 70B Turbo

Llama 3.1 · 70.6B

GPU
H100 SXM
Time
4.1h
Cost
$7.7
Quality
88%
Fine-tune this model
#14NEFTune

GoogleGemma 3 1B

Gemma 3 · 1B

GPU
T4
Time
53m
Cost
$0.11
Quality
85%
Fine-tune this model
#15QLoRA

MistralBioMistral 7B

BioMistral · 7.2B

GPU
T4
Time
6.3h
Cost
$0.76
Quality
88%
Fine-tune this model
#16QLoRA

MistralNV EmbedQA Mistral 7B

NV EmbedQA · 7.2B

GPU
T4
Time
6.3h
Cost
$0.76
Quality
88%
Fine-tune this model
#17QLoRA

MistralMistral 7B

Mistral · 7.3B

GPU
T4
Time
6.4h
Cost
$0.77
Quality
88%
Fine-tune this model
#18NEFTune

MetaTinyLlama 1.1B Chat

TinyLlama · 1.1B

GPU
T4
Time
58m
Cost
$0.12
Quality
85%
Fine-tune this model
#19NEFTune

MetaTinyLlama 1.1B

TinyLlama · 1.1B

GPU
T4
Time
58m
Cost
$0.12
Quality
85%
Fine-tune this model
#20QLoRA

AlibabaQwen 2.5 14B

Qwen 2.5 · 14.8B

GPU
T4
Time
13h
Cost
$1.6
Quality
88%
Fine-tune this model
#21QLoRA

AlibabaQwen 2.5 72B

Qwen 2.5 · 72.7B

GPU
H100 SXM
Time
4.2h
Cost
$7.9
Quality
88%
Fine-tune this model
#22GaLore

MicrosoftPhi 2

Phi · 2.7B

GPU
T4
Time
2.4h
Cost
$0.28
Quality
97%
Fine-tune this model
#23ReFT

AlibabaQwen 2.5 32B

Qwen 2.5 · 32.5B

GPU
H100 SXM
Time
1.9h
Cost
$3.5
Quality
83%
Fine-tune this model
#24GaLore

DeepSeekDeepSeek R1 Distill 1.5B

DeepSeek R1 · 1.5B

GPU
T4
Time
1.3h
Cost
$0.16
Quality
97%
Fine-tune this model
#25QLoRA

DeepSeekDeepSeek R1 Distill 8B

DeepSeek R1 · 8B

GPU
T4
Time
7.0h
Cost
$0.84
Quality
88%
Fine-tune this model
#26QLoRA

DeepSeekDeepSeek R1 Distill 14B

DeepSeek R1 · 14.8B

GPU
T4
Time
13h
Cost
$1.6
Quality
88%
Fine-tune this model
#27ReFT

DeepSeekDeepSeek R1 Distill 32B

DeepSeek R1 · 32.8B

GPU
H100 SXM
Time
1.9h
Cost
$3.6
Quality
83%
Fine-tune this model
#28QLoRA

DeepSeekDeepSeek R1 Distill 70B

DeepSeek R1 · 70.6B

GPU
H100 SXM
Time
4.1h
Cost
$7.7
Quality
88%
Fine-tune this model
#29QLoRA

DeepSeekDeepSeek R1

DeepSeek R1 · 671B MoE (37B active)

GPU
T4x31
Time
24h
Cost
$91
Quality
88%
Fine-tune this model
#30QLoRA

DeepSeekDeepSeek V3

DeepSeek V3 · 671B MoE (37B active)

GPU
T4x31
Time
24h
Cost
$91
Quality
88%
Fine-tune this model
#31NEFTune

DeepSeekDeepSeek V3-0324

DeepSeek V3 · 685B MoE (37B active)

GPU
T4x564
Time
1.4h
Cost
$92
Quality
85%
Fine-tune this model
#32QLoRA

MetaLlama 3 70B 1M Context

Llama 3 · 70.6B

GPU
H100 SXM
Time
4.1h
Cost
$7.7
Quality
88%
Fine-tune this model
#33QLoRA

MetaLlama 3 70B

Llama 3 · 70.6B

GPU
H100 SXM
Time
4.1h
Cost
$7.7
Quality
88%
Fine-tune this model
#34QLoRA

MistralMixtral 8x7B Instruct

Mixtral · 46.7B MoE (12.9B active)

GPU
H100 SXM
Time
2.7h
Cost
$5.1
Quality
88%
Fine-tune this model
#35QLoRA

MistralMixtral 8x7B

Mixtral · 46.7B MoE (12.9B active)

GPU
H100 SXM
Time
2.7h
Cost
$5.1
Quality
88%
Fine-tune this model
#36QLoRA

GoogleGemma 2 9B

Gemma 2 · 9.2B

GPU
T4
Time
8.1h
Cost
$0.97
Quality
88%
Fine-tune this model
#37ReFT

MicrosoftPhi 4 Mini

Phi · 3.8B

GPU
T4
Time
3.3h
Cost
$0.40
Quality
83%
Fine-tune this model
#38QLoRA

MicrosoftPhi-4

Phi · 14.7B

GPU
T4
Time
13h
Cost
$1.5
Quality
88%
Fine-tune this model
#39QLoRA

MetaLlama 2 7B

Llama 2 · 7B

GPU
T4
Time
6.1h
Cost
$0.74
Quality
88%
Fine-tune this model
#40QLoRA

MetaLlama 2 13B

Llama 2 · 13B

GPU
T4
Time
11h
Cost
$1.4
Quality
88%
Fine-tune this model
#41QLoRA

MetaLlama 2 70B

Llama 2 · 70B

GPU
H100 SXM
Time
4.0h
Cost
$7.6
Quality
88%
Fine-tune this model
#42NEFTune

MetaLlama Guard 3 1B

Llama Guard · 1B

GPU
T4
Time
53m
Cost
$0.11
Quality
85%
Fine-tune this model
#43QLoRA

MetaCode Llama 7B

Code Llama · 7B

GPU
T4
Time
6.1h
Cost
$0.74
Quality
88%
Fine-tune this model
#44QLoRA

MetaLlama 3.3 8B

Llama 3.3 · 8B

GPU
T4
Time
7.0h
Cost
$0.84
Quality
88%
Fine-tune this model
#45QLoRA

MetaLlama Guard 3 8B

Llama Guard · 8B

GPU
T4
Time
7.0h
Cost
$0.84
Quality
88%
Fine-tune this model
#46QLoRA

MetammE5-mllama-11b-instruct

intfloat · 10.6B

GPU
T4
Time
9.3h
Cost
$1.1
Quality
88%
Fine-tune this model
#47QLoRA

MetaLlama 3.2 11B Vision

Llama 3.2 · 11B

GPU
T4
Time
9.6h
Cost
$1.2
Quality
88%
Fine-tune this model
#48QLoRA

MetaCode Llama 13B

Code Llama · 13B

GPU
T4
Time
11h
Cost
$1.4
Quality
88%
Fine-tune this model
#49ReFT

MetaCode Llama 34B

Code Llama · 34B

GPU
H100 SXM
Time
2.0h
Cost
$3.7
Quality
83%
Fine-tune this model
#50QLoRA

MetaLlama 3.1 Nemotron 51B

Llama 3.1 · 51B

GPU
H100 SXM
Time
2.9h
Cost
$5.5
Quality
88%
Fine-tune this model
Showing 1–50 of 311 modelsPer page
Page 1 of 7

Ready to fine-tune your model?

Use the training calculator to get detailed cost estimates for any model, method, and GPU combination.

Open Training Calculator