Meta Llama: Llama 3.2 3B Instruct
Try Llama 3.2 3B Instruct for free in your browser, and add it to your app for free with Puter.js AI API.
Try it free Add to your app// npm install @heyputer/puter.js
import { puter } from '@heyputer/puter.js';
puter.ai.chat("Explain quantum computing in simple terms", {
model: "meta-llama/llama-3.2-3b-instruct"
}).then(response => {
document.body.innerHTML = response.message.content;
});
<html>
<body>
<script src="https://js.puter.com/v2/"></script>
<script>
puter.ai.chat("Explain quantum computing in simple terms", {
model: "meta-llama/llama-3.2-3b-instruct"
}).then(response => {
document.body.innerHTML = response.message.content;
});
</script>
</body>
</html>
Model Card
Llama 3.2 3B Instruct is a compact 3 billion parameter model optimized for on-device use cases with 128K context support. It outperforms comparable models on instruction following, summarization, and tool-use tasks.
Context Window 131K
tokens
Max Output 72K
tokens
Input Cost $0.05
per million tokens
Output Cost $0.34
per million tokens
Input text
modalities
Tool Use No
Release Date Sep 25, 2024
Try Llama 3.2 3B Instruct for free
Try Llama 3.2 3B Instruct instantly in your browser.
This playground uses the Puter.js AI API — no API keys or setup required.
Benchmarks
How Llama 3.2 3B Instruct performs on standard evaluations.
| Benchmark | Score |
|---|---|
| GPQA Diamond Graduate-level science Q&A | 25.5% |
| Humanity's Last Exam Cross-domain reasoning | 5.3% |
| LiveCodeBench Recent coding problems | 8.3% |
| MATH-500 Competition math | 48.9% |
| AIME 2024 Advanced math exam | 6.7% |
| AIME 2025 Advanced math exam | 3.3% |
| IFBench Instruction following | 26.2% |
| LCR Long-context reasoning | 4.3% |
| τ²-Bench Tool use / agents | 21.1% |
Scores sourced from Artificial Analysis.
Find other Meta Llama models →
Muse Spark Open
Muse Spark Open is an open-weight multimodal model from Meta, released on October 15, 2026. It is a mixture-of-experts model with 816B total and 42B active parameters and a 1,048,576-token context window. It accepts text, image, video, and audio input and returns text. It is aimed at agentic and tool-use work. Meta reports that it leads the open models it compared against on the agentic benchmarks MCP-Atlas and JobBench, and that it trails some closed Chinese models such as Kimi K3 and GLM 5.3 on the coding benchmark Terminal-Bench 2.1. These results are vendor-reported. It supports tool calling, structured output, and configurable reasoning effort. It suits developers who want an agent-capable model with open weights from a US lab.
ChatMuse Wonder
Muse Wonder is a multimodal agent model from Meta Superintelligence Labs, built on Muse Spark and released on October 12, 2026. It accepts text, images, video, audio, and PDF input, returns text, and has a 1,048,576-token context window. It is designed for long, multi-step agent work such as coding, legal and financial documents, and tasks across several apps. It supports tool calling, structured output, and configurable reasoning effort. Meta reports 78.8 on DeepSWE 1.1 and 53.1 on SWE-Marathon, and Artificial Analysis gives it an Intelligence Index score of 52, up from 48 for Muse Spark 1.3. Its per-token price is below Claude Fable 5.1 ($10 input, $50 output), though Claude Opus 5.5 scores higher on the index and on OSWorld.
ChatMuse Spark 1.3
Muse Spark 1.3 is Meta Superintelligence Labs' follow-up to the coding-focused Muse Spark 1.2, released on September 2, 2026. It targets long-horizon agentic work, multi-agent orchestration, and extended multi-step coding tasks. On Meta's own harness it scored 88.8% on Terminal-Bench 2.1, level with OpenAI's GPT-5.6 Sol, and 75.4% on DeepSWE 1.1, up from 59.3% for 1.2. These are vendor-reported numbers, not yet independently verified. Artificial Analysis gives it an Intelligence Index score of 48, up from 40 for Muse Spark 1.2, and Anthropic's Opus 5 still leads it on several agent benchmarks. It keeps the 1,048,576-token context window, text, image, video, and PDF input, and tool calling from earlier versions, at the same $1.25 per million input and $4.25 per million output token pricing.
Frequently Asked Questions
You can access Llama 3.2 3B Instruct by Meta Llama through Puter.js AI API. Include the library in your web app or Node.js project and start making calls with just a few lines of JavaScript — no backend and no configuration required. You can also use it with Python or cURL via Puter's OpenAI-compatible API.
Llama 3.2 3B Instruct is free to try with a Puter account. Every account includes a free AI allowance, and you can chat with it in the playground on this page. You can upgrade your account anytime for a larger allowance.
Llama 3.2 3B Instruct is free to integrate using the Puter.js AI API. With the User-Pays Model, you can add AI to your app for $0, since users cover their own AI usage through their Puter account.
| Price per 1M tokens | |
|---|---|
| Input | $0.05 |
| Output | $0.34 |
Llama 3.2 3B Instruct was created by Meta Llama and released on Sep 25, 2024.
Llama 3.2 3B Instruct supports a context window of 131K tokens. For reference, that is roughly equivalent to 262 pages of text.
Llama 3.2 3B Instruct can generate up to 72K tokens in a single response.
Llama 3.2 3B Instruct accepts the following input types: text. It produces: text.
No, Llama 3.2 3B Instruct does not currently support tool use (function calling).
Llama 3.2 3B Instruct scores 5.7 on the Artificial Analysis Intelligence Index, outperforming 10% of tracked models. On math, it scores 3.3 (outperforms 4% of models).
Yes — the Llama 3.2 3B Instruct API works with any JavaScript framework, Node.js, or plain HTML through Puter.js. Just include the library and start building. See the documentation for more details.
Add Llama 3.2 3B Instruct to your app for free
Developers can integrate Llama 3.2 3B Instruct for free using the Puter.js AI API.
With the User-Pays Model, each user covers their own AI usage instead of the developer.