Llama 3.3 70B

Llama 3.3 70B was Meta’s answer to an expensive problem: it comes close to the much larger Llama 3.1 405B on key tests at about a sixth of the size. It is a plain, predictable model — no reasoning step, so answers start quickly — that follows instructions very well (92.1% on IFEval) and writes clean code for its class (88.4% on HumanEval).

It has been around since December 2024, so it is well understood and widely tested. That makes it a steady choice for drafting, summarising and answering from your own documents, where consistent behaviour matters more than the newest benchmark score.

For scale: 1,000 requests of about 3,000 tokens in and 1,000 out cost $4.07 — a long prompt and a page of answer each. Every rate is published and stays the same from one request to the next. See every model.

Llama 3.3 70B is billed in rupees, on the same monthly invoice as every other model your team uses — no foreign card, no separate Meta account, and one key for all of it.

Prompts sent to Llama 3.3 70B are processed outside Sri Lanka. If a workload has to stay in the country, Qwen3.8-27B (Colombo) runs on our own servers in Colombo.

Meta’s supported languages for Llama 3.3 include Hindi and Thai but not Sinhala or Tamil. Try it on your own documents before relying on it for either.

Use the OpenAI SDK you already have — point it at Roar AI and name the model. Read the docs.

from openai import OpenAI

client = OpenAI(base_url="https://api.roar-ai.com/v1", api_key="roar_live_…")

reply = client.chat.completions.create(
    model="llama-3.3-70b",
    messages=[{"role": "user", "content": "Summarise this contract in plain English."}],
)