Meta (Llama)
Llama 3.3 70B
model
openrouter/llama-3.3-70bBudgetStableStreamingToolsBatchLong contextUnited States
Context
131K
Input / 1M tokens
$0.1000
Output / 1M tokens
$0.3200
About this model
Llama 3.3 70B
Llama 3.3 70B is Meta's late-2024 dense flagship — comparable quality to Llama 3.1 405B at a sixth of the inference cost. Strong general-purpose model with high throughput.
Released
2024-12
Training cutoff
2023-12
Parameters
70B (dense)
Best for
Use cases
- ✓Production chat replacing GPT-4-class workloads
- ✓RAG over long documents
- ✓Multi-lingual customer support
- ✓Synthetic data generation at scale
Strengths
What it does well
- ●Near-frontier quality at open-weights pricing
- ●128K context window
- ●Strong tool calling + JSON mode
- ●Permissive license
Limitations
Know the trade-offs
- !No native vision — pair with a VL model if multimodal
- !Reasoning depth trails dedicated reasoners (R1, o1)
- !Knowledge cutoff December 2023
Benchmarks
Published scores
Scores from official model cards and public leaderboards. Higher is better unless noted.
| Benchmark | Score | Measures |
|---|---|---|
| MMLU | 86.0 | General knowledge across 57 subjects |
| HumanEval | 88.4 | Python code generation, pass@1 |
| MATH | 77.0 | Mathematics, mixed difficulty |
| IFEval | 92.1 | Instruction following |
Code example
Chat completion with Llama 3.3 70B
from openai import OpenAI
client = OpenAI(
base_url="https://api.anoman.io/v1",
api_key="anm-sk-..."
)
response = client.chat.completions.create(
model="openrouter/llama-3.3-70b",
messages=[{"role": "user", "content": "Hello!"}]
)
print(response.choices[0].message.content)Alternatives
Similar models in our catalog
Workloads that fit Llama 3.3 70B often fit these too. Compare benchmarks and pricing before locking in a default.