Llama 3 Models by AI at Meta
Details
Capabilities
About
Llama 3, developed by Meta AI and released in April 2024, represents a significant advancement in large language models (LLMs). Available in two configurations—8 billion and 70 billion parameters—the models offer both pretrained and instruction-tuned versions, enhancing their adaptability and effectiveness in dialogue scenarios. Llama 3 sets itself apart by being trained on over 15 trillion tokens of publicly available data, a massive expansion over its predecessor, Llama 2, and includes a substantial increase in code data. The models not only excel in performance but also incorporate robust safety features like Llama Guard 2 and Code Shield, underscoring Meta's focus on responsible AI use. Llama 3 models are accessible on platforms such as AWS, Google Cloud, and Hugging Face, with plans for future updates that will expand their capabilities to include multimodal functionalities and multilingual support.
Current Variants
Use-when guidance is derived from seed capabilities, context, release, and replacement fields.
Use when the workload needs 8k context, 8B parameters, and tool use.
Use when the workload needs 8k context and 70B parameters.
Use when the workload needs 8k context, 70B parameters, and structured outputs.
Use when the workload needs 8k context, 8B parameters, and structured outputs.
Use when the workload needs 8k context and 70B parameters.
Use when the workload needs 8k context and 8B parameters.
Use when the workload needs 8k context, 8B parameters, and structured outputs.
Use when the workload needs 8k context, 70B parameters, and structured outputs.
Use when the workload needs 8k context, 8B parameters, and structured outputs.
Use when the workload needs 8k context, 70B parameters, and structured outputs.
Use when the workload needs 8k context and 8B parameters.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| Together AI - Llama 3 8B Lite | Use when the workload needs 8k context, 8B parameters, and tool use. | 2025-07 | 8k context8B parameterstool use | Current |
| Llama 3 Taiwan 70B Instruct | Use when the workload needs 8k context and 70B parameters. | 2024-07 | 8k context70B parameters | Current |
| Llama 3 70B Instruct | Use when the workload needs 8k context, 70B parameters, and structured outputs. | 2024-04 | 8k context70B parametersstructured outputs | Current |
| Llama 3 8B Instruct | Use when the workload needs 8k context, 8B parameters, and structured outputs. | 2024-04 | 8k context8B parametersstructured outputs | Current |
| Llama 3 70B | Use when the workload needs 8k context and 70B parameters. | 2024-04 | 8k context70B parameters | Current |
| Llama 3 8B | Use when the workload needs 8k context and 8B parameters. | 2024-04 | 8k context8B parameters | Current |
| Together AI Llama-3-8B-Instruct | Use when the workload needs 8k context, 8B parameters, and structured outputs. | 2024-04 | 8k context8B parametersstructured outputs | Current |
| Together AI Llama-3-70B-Instruct | Use when the workload needs 8k context, 70B parameters, and structured outputs. | 2024-04 | 8k context70B parametersstructured outputs | Current |
| DeepInfra Llama 3 8B Instruct | Use when the workload needs 8k context, 8B parameters, and structured outputs. | 2024-04 | 8k context8B parametersstructured outputs | Current |
| DeepInfra Llama 3 70B Instruct | Use when the workload needs 8k context, 70B parameters, and structured outputs. | 2024-04 | 8k context70B parametersstructured outputs | Current |
| Fireworks Llama-3-8B-Instruct | Use when the workload needs 8k context and 8B parameters. | 2024-04 | 8k context8B parameters | Current |
Release Timeline
3 release groupsSpecifications(11 models)
| Model | Released | Context | Parameters | Fn Calling | Tool Use | Structured Outputs |
|---|---|---|---|---|---|---|
| Together AI - Llama 3 8B Lite | 2025-07 | 8k | 8B | Yes | Yes | Yes |
| Llama 3 Taiwan 70B Instruct | 2024-07 | 8k | 70B | No | No | No |
| Llama 3 70B Instruct | 2024-04 | 8k | 70B | No | No | Yes |
| Llama 3 8B Instruct | 2024-04 | 8k | 8B | No | No | Yes |
| Llama 3 70B | 2024-04 | 8k | 70B | No | No | No |
| Llama 3 8B | 2024-04 | 8k | 8B | No | No | No |
| Together AI Llama-3-8B-Instruct | 2024-04 | 8k | 8B | No | No | Yes |
| Together AI Llama-3-70B-Instruct | 2024-04 | 8k | 70B | No | No | Yes |
| DeepInfra Llama 3 8B Instruct | 2024-04 | 8k | 8B | No | No | Yes |
| DeepInfra Llama 3 70B Instruct | 2024-04 | 8k | 70B | No | No | Yes |
| Fireworks Llama-3-8B-Instruct | 2024-04 | 8k | 8B | No | No | No |
Available From(20 providers)
Pricing
Frequently Asked Questions
- What is Llama 3 used for?
- Llama 3 is used for agent workflows and tool use, structured outputs, and coding. The family description and listed model capabilities point to those workloads as the best fit.
- How does Llama 3 compare to MOSS-Audio?
- Llama 3 by AI at Meta is strongest where you need agent workflows and tool use, while MOSS-Audio by MOSI AI is the closest related family to check for multimodal. Llama 3 has 11 listed variants and reaches up to 8k context, so compare the specs and pricing tables before choosing a production model.
- Which Llama 3 model should I use?
- For the lowest listed input price, start with Llama 3 8B Instruct through OpenRouter at $0.03/1M input tokens. For the most capable/latest local choice, evaluate Together AI - Llama 3 8B Lite with 8k context and tool use, function calling, and structured outputs.
Related Model Families
Models(11)
Together AI - Llama 3 8B Lite
Llama 3 Taiwan 70B Instruct
Llama 3 70B Instruct
Llama 3 8B Instruct
Llama 3 70B
Llama 3 8B
Together AI Llama-3-8B-Instruct
Together AI Llama-3-70B-Instruct
DeepInfra Llama 3 8B Instruct
DeepInfra Llama 3 70B Instruct
Fireworks Llama-3-8B-Instruct