News
ReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior ReleaseReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior Release
EveryLLM
fnLlama 3.3 70B

Llama 3.3 70B

A text-only 70B instruction-tuned model from Meta's Llama 3 line.

Llama 3.3 70B is an instruction-tuned large language model released by Meta as part of the Llama 3 family. The company says it improves on Llama 3.1 70B and, for text-only work, on Llama 3.2 90B, and that it can approach Llama 3.1 405B for some uses.

Share on WhatsApp

  • Text-only 70B instruction-tuned model in the Llama 3 family
  • Meta says it beats Llama 3.1 70B and, for text, Llama 3.2 90B
  • Company claims it can approach Llama 3.1 405B on some applications
  • No pretrained base version is released
  • Prompts from Llama 3.1 run unchanged; tools use the Llama 3.2 format

What it is

Llama 3.3 70B is a text-only, instruction-tuned language model from Meta's Llama 3 family. It has 70 billion parameters and is aimed at text applications. Meta says it performs better than Llama 3.1 70B, and better than Llama 3.2 90B when the task is text only. For some uses, the company says it comes close to the much larger Llama 3.1 405B. Only the instruction-tuned form is released; there is no pretrained base version.

How it came about

Llama 3.3 sits inside the Llama 3 line, after Llama 3.1 and Llama 3.2. It reuses the prompt format of Llama 3.1, so prompts written for that version run unchanged. Its function-calling format follows Llama 3.2 rather than 3.1. Meta points to an official model card on GitHub for full technical detail.

Inside the model

The model is instruction-tuned, meaning it was further trained to follow directions rather than just predict text. It keeps the Llama 3.1 prompt layout, with system, user and assistant turns marked by special tokens. For tools, it supports the Llama 3.2 function-calling style, which is more flexible than the 3.1 version: the list of available functions can sit in the user message as well as the system message. The model can name several calls at once, including calls in parallel, though it only writes the calls and does not execute them.

What it can do

Meta reports gains over Llama 3.1 70B and, for text-only tasks, over Llama 3.2 90B, and says performance can approach Llama 3.1 405B in some applications. It handles zero-shot function calling, returning calls in a set format, and works with built-in tools: Brave Search for web queries, Wolfram Alpha for maths, and a code interpreter that emits Python. No benchmark numbers are given here.

How to use it

The model is distributed as an instruction-tuned release only. It follows the Llama 3.1 prompt format and the Llama 3.2 tool-calling format, and documentation lives on Meta's Llama site and GitHub model card.

Questions people ask about Llama 3.3 70B

What is Llama 3.3 70B?

It is a text-only, instruction-tuned language model with 70 billion parameters from Meta, part of the Llama 3 family. It is meant for text tasks and for calling tools such as search, maths and a Python code interpreter.

How does it compare with other Llama models?

Meta says it improves on Llama 3.1 70B and, for text-only work, on Llama 3.2 90B. The company also says that for some applications it gets close to the much larger Llama 3.1 405B.

Is there a base or pretrained version?

No. Llama 3.3 70B is supplied only in an instruction-tuned form. A pretrained version is not available.

Do my old Llama prompts still work?

Yes. Llama 3.3 uses the same prompt format as Llama 3.1, so prompts written for that version work unchanged. Function calling follows the newer Llama 3.2 format instead.

Good reading and sources

    Where this page comes from

    There is no English Wikipedia article on Llama 3.3 70B yet, so the facts on this page come from the sources listed above: llama.com/docs/model-cards-and-prompt-formats/llama3_3/. Some are the subject's own pages, so its claims about itself are reported as its claims. EveryLLM wrote the text in its own words, with AI assistance, and compares it with the sources to make sure no passage is copied.

    It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.

    The weekly digest

    Every Monday: the model of the day, new pages on EveryLLM, and the conferences and anniversaries of the week ahead.