Announcing Llama 3.1 and GPT-4o Mini Fine-Tuning on FinetuneDB

Introducing Llama 3.1 and GPT-4o Mini

We’re excited to introduce three new models now available on FinetuneDB: Llama 3.1 8B and 70B, and GPT-4o Mini.

These models bring significant improvements in processing speed, response accuracy, and efficiency for AI applications.

Llama 3.1 (8B and 70B) and GPT-4o Mini

Model Quality and Performance

Llama 3.1 70B

Context: 128K, Speed: Medium

With a significantly larger parameter set, 70B excels in dealing with extensive datasets, producing highly sophisticated and contextually rich responses. Its amplified context window allows it to consider a broader range of information, making it ideal for in-depth analyses and detailed content creation.

Llama 3.1 8B

Context: 128K, Speed: Fastest

Has been refined to tackle complex tasks more efficiently than its predecessors. It has improved in handling multi-step tasks with better alignment and response diversity. It is particularly suited for natural language understanding and generation, maintaining high accuracy and nuanced outputs across different scenarios.

GPT-4o Mini

Context: 128K, Speed: Fast

Designed for efficiency, this model offers robust performance in resource-constrained environments. It balances power and computational efficiency, making it suitable for applications like mobile devices and real-time data processing where resources are limited but high performance is required.

Pricing Details

These models are now available for inference on FinetuneDb at a per-token cost:

Llama 3.1 70B:

Training Tokens: $6 / 1M, Input and Output Tokens: $1.10 / 1M

Llama 3.1 8B:

Training Tokens: $2 / 1M, Input and Output Tokens: $0.30 / 1M

GPT-4o Mini:

Training Tokens: $3 / 1M, Input and Output Tokens: $0.30 / 1M, Output Tokens: $1.20 / 1M

Fine-tuning and serving on FinetuneDB

Use your datasets, fine-tune these new models, and deploy them straight from our FintuneDB. We supports the entire process, from dataset management to model training and serving with an end-to-end workflow.

For more details, visit our documentation to get started, or reach out to us with any questions.