Llama 3.3 70B Instruct
Open Llama instruction model for multilingual chat, reasoning, and coding
Cheapest input
$0.05
$/MTok · models.dev
Cheapest output
$0.23
$/MTok
Context
131K
tokens
Offers
16
3 sources
Offers
Same model, different listings. First-party, router, and aggregator rows stay separate so you can see the spread instead of a single blended number.
| Source | Provider | Input | Output | Cache read | Context |
|---|---|---|---|---|---|
| models.devcheapestmeta-llama/llama-3.3-70b-instruct | NanoGPT | $0.05 | $0.23 | $0 | 131K |
| models.devmeta-llama/Llama-3.3-70B-Instruct | Meganova | $0.1 | $0.3 | $0 | 131K |
| models.devmeta-llama/llama-3.3-70b-instruct | Kilo Gateway | $0.1 | $0.32 | $0 | 131K |
| models.devmeta-llama/Llama-3.3-70B-Instruct | IO.NET | $0.13 | $0.38 | $0.065 | 128K |
| models.devmeta-llama/Llama-3.3-70B-Instruct | Nebius | $0.13 | $0.4 | $0.013 | 128K |
| OpenRoutermeta-llama/llama-3.3-70b-instruct | OpenRouter | $0.13 | $0.4 | $0 | 131K |
| models.devmeta-llama/llama-3.3-70b-instruct | Novita AI | $0.135 | $0.4 | $0 | 131K |
| LiteLLMmeta-llama/llama-3.3-70b-instruct | — | $0.135 | $0.4 | $0.0135 | — |
| LiteLLMmeta-llama/Llama-3.3-70B-Instruct | — | $0.23 | $0.4 | $0.023 | — |
| models.devmeta-llama/Llama-3.3-70B-Instruct | Hugging Face | $0.59 | $0.79 | $0 | 131K |
| models.devmeta-llama/Llama-3.3-70B-Instruct | Weights & Biases | $0.71 | $0.71 | $0.71 | 128K |
| LiteLLMmeta-llama/llama-3-3-70b-instruct | — | $0.71 | $0.71 | $0.071 | — |
| models.devmeta-llama/Llama-3.3-70B-Instruct | Pioneer | $0.9 | $0.9 | $0.9 | 131K |
| models.devmeta-llama/Llama-3.3-70B-Instruct | Berget.AI | $0.99 | $0.99 | $0 | 128K |
| models.devmeta-llama/Llama-3.3-70B-Instruct | CloudFerro Sherlock | $2.92 | $2.92 | $0 | 70K |
| LiteLLMwandb/meta-llama/Llama-3.3-70B-Instruct | — | $71000 | $71000 | $7100 | — |
Capabilities
- Reasoning
- Tool calling
- Structured output
- Attachments
- Open weights
- In: text
- Out: text
Knowledge cutoff: 2023-12Released: 2025-04-27Family: llama
Scores
Sparse coverage on purpose. Missing a score means we have not linked one yet, not that the model is unranked. Full boards on /benchmarks.
- AA intelligence9.4
- AA coding11.9
- AA agentic0.3
Sources: Arena AI, Artificial Analysis. See /benchmarks.