Meta
Llama 3.1 70B
Open Llama instruction model for multilingual chat, reasoning, and coding
Model facts
- Context window
- 128K tokens
- Maximum output
- 16K tokens
- Cheapest paid input
- $0.4 per 1M tokens
- Cheapest paid output
- $0.4 per 1M tokens
- Open weights
- yes
- Providers
- 7
Capabilities and modalities
tool calling, structured output, temperature control, open weights, text.
Local hardware estimate
At 4K context: Q4_K_M 44.2 GiB, Q8_0 77.0 GiB, F16 138.6 GiB working memory.
- Parameters
- 70.6 billion
- Architecture
- llama
- Model license
- llama3.1
Planning estimate adapted from the MIT-licensed whichllm estimator using metadata from Hugging Face; not a vendor minimum requirement.
Artificial Analysis benchmarks
Compare this model on the full LLM benchmark leaderboard.
| Configuration | AAI | Coding | Math | Output tokens/s |
|---|---|---|---|---|
| default | 6.5 | not measured | 4.0 | not measured |
Observed price history
- 2026-08-24: $0.4 input / $0.4 output per 1M tokens
API providers
- Nvidia — model id meta/llama-3.1-70b-instruct; free input / free output per 1M tokens; provider documentation
- OpenRouter — model id meta-llama/llama-3.1-70b-instruct; $0.4 input / $0.4 output per 1M tokens; provider documentation
- Kilo Gateway — model id meta-llama/llama-3.1-70b-instruct; $0.4 input / $0.4 output per 1M tokens; provider documentation
- Amazon Bedrock — model id meta.llama3-1-70b-instruct-v1:0; $0.72 input / $0.72 output per 1M tokens; provider documentation
- DevPass (LLM Gateway) — model id llama-3.1-70b-instruct; $0.72 input / $0.72 output per 1M tokens; provider documentation
- Vercel AI Gateway — model id meta/llama-3.1-70b; $0.72 input / $0.72 output per 1M tokens; provider documentation
- Weights & Biases — model id meta-llama/Llama-3.1-70B-Instruct; $0.8 input / $0.8 output per 1M tokens; provider documentation