Llama 3.3 70b specs, prices, where it runs and its benchmark standing
open weights · top 10 on 0 of 2 · released Dec 6, 2024
Facts
- Provider
- Meta
- Hosting
- 12 priced API routes · Azure OpenAI (31 regions) · OpenRouter · self-host: weights on Hugging Face
- PricingRegions
- Hardware
- Estimate: about 141 GB of weights as published (BF16), about 71 GB at 8-bit, about 35 GB at 4-bit, plus memory for the context
- Hugging Face config
- Weights
- Open weights
- Benchmarks
- 2 boards captured
| Rank | Board | Score | |||
|---|---|---|---|---|---|
| 44 | Agents and tools | 31.9% | 81 | ||
| 100 | Reasoning | 65.9% | 100 |
Versions and release history
6 versions in this line since Jul 16, 2024, from Pricing, OpenRouter and the boards.
| Version | Released | |
|---|---|---|
| Llama 3.3 70Bthis model | Dec 6, 20241 year ago | Pricing |
| Llama 3.2 1B | Sep 25, 20242 years ago | Pricing |
| Llama 3.2 3B | Sep 25, 20242 years ago | Pricing |
| Llama 3.1 70B Instruct | Jul 23, 20242 years ago | Pricing |
| Llama 3.1 8B | Jul 23, 20242 years ago | Pricing |
| Llama 3.1 405B | Jul 16, 20242 years ago | Hugging Face |
Category ranks, more models and catalog changes
Rank in each category
More from Meta
Catalog changes, newest first (kept for good)
- no longer listed on AWS BedrockNov 2, 2026 7 days ago
- now on Azure OpenAIOct 26, 2026 2 weeks ago