LLaMA

Top Builders

Explore the top contributors showcasing the highest number of app submissions within our community.

LLaMA (Large Language Model Meta AI)

LLaMA is a state-of-the-art foundational large language model designed to help researchers advance their work in the subfield of AI. It is available in multiple sizes (7B, 13B, 33B, and 65B parameters) and aims to democratize access to large language models by requiring less computing power and resources for training and deployment. LLaMA is developed by the FAIR team of Meta AI and has been trained on a large set of unlabeled data, making it ideal for fine-tuning for a variety of tasks.

General
Release date	2023
Author	Meta AI FAIR Team
Model sizes	7B, 13B, 33B, 65B parameters
Model Architecture	Transformer
Training data source	CCNet, C4, GitHub, Wikipedia, Books, ArXiv, Stack Exchange
Supported languages	20 languages with Latin and Cyrillic alphabets

Start building with LLaMA

LLaMA provides an opportunity for researchers and developers to study large language models and explore their applications in various domains. To get started with LLaMA, you can access its code through the GitHub repository.

LLaMA Links

Important links about LLaMA in one place:

LLaMA Research Paper - Read the research paper about LLaMA to learn more about the model and its development process.
LLaMA GitHub Repository - Access the LLaMA model, code, and resources on GitHub.

Edit on GitHub

Meta LLaMA AI technology Hackathon projects

Discover innovative solutions crafted with Meta LLaMA AI technology, developed by our community members during our engaging hackathons.

Archie

AI agents are becoming incredibly capable at reasoning and coding. Yet a hard limit remains. The moment a task requires payment, the agent stops. This is because the internet is built for humans. Payments require credit cards and manual accounts. This makes software dependent on human intervention. Archie represents a shift toward machine native commerce. We believe that autonomous software should be able to spend money per action as easily as it spends compute per token. Archie is an agent that manages its own wallet. It uses Circle Nanopayments to buy premium data in real time. Our system uses the Arc Network to settle sub cent transactions. We use the x402 protocol to bridge the gap between intelligence and value. Every action is recorded on the blockchain as a proof of reputation. This creates a verifiable history of reliability for every agent. This is not just a tool. It is an economic participant. Archie demonstrates that when transaction costs disappear, agents can finally pay their own way. We are building the infrastructure for a world where machines transact directly with one another without human oversight.

InferPay: Pay-Per-Inference for the Agent Economy

InferPay is payment infrastructure for the agent economy on Arc Testnet. It allows AI agents to access LLMs and pay in USDC for each request using x402 and Circle Nanopayments, without API keys, user accounts, or subscription billing. The product is built around a simple idea: if agents are becoming economic actors, they need a native way to buy inference and services on the internet. InferPay provides that flow by combining pay-per-request access, machine-readable payment negotiation, and micropayment settlement into one system. The current version already includes a working x402 proxy, buyer agent flow, playground, Circle Nanopayments integration, and dynamic pricing, routing, and usage visibility. Rather than forcing a premature marketplace, this project focuses first on the core payment primitive that agent-to-agent commerce depends on: an agent can request inference, pay in USDC, and receive a result in one machine-native flow. The goal of this submission is to show that machine-to-machine AI payments are not just a concept, but a working infrastructure layer for agentic applications.

NeuroPay — Pay-Per-Inference AI on Arc

NeuroPay is a live, deployed Pay-Per-Inference AI API competing in the Usage-Based Compute Billing track. It is hosted at https://neuropay-arc.onrender.com and powered end-to-end by Circle's infrastructure on Arc. THE PROBLEM: AI inference requires fair usage-based pricing — but traditional blockchain gas fees destroy the economics. Ethereum gas costs $2.50 per transaction, which is 1,667 times more than a $0.0015 inference charge. Even Polygon at $0.05 gas is 33 times the payment value. Sub-cent AI billing is economically impossible on traditional chains. THE SOLUTION: Circle Nanopayments on Arc eliminates gas overhead entirely. NeuroPay charges exactly $0.001 per 100 tokens in USDC, settled on Arc in under one second with zero gas cost — giving providers 100% margin retention. WHAT WE BUILT: - Real AI inference using Groq's Llama 3.1 model — actual tokens counted, actual billing - Circle Nanopayments integration for gas-free USDC settlement on Arc - Agent-to-agent payment simulation showing autonomous AI economics - Live transaction feed with Arc tx hashes verifiable on the block explorer - Cumulative analytics dashboard tracking USDC volume vs traditional gas costs - Full REST API with /infer, /deposit, /stats, /margin-analysis endpoints - Professional web dashboard showing real-time billing metrics KEY PROOF POINTS: - 55+ on-chain transactions demonstrated at sub-cent pricing - $0.00 gas cost across all transactions via Nanopayments - Traditional gas equivalent saved: $5.50+ across the demo - Margin multiplier: 66x more efficient than Polygon, 1,667x vs Ethereum NeuroPay proves that usage-based AI billing at machine scale is only viable on Arc with Circle Nanopayments — and shows exactly what the agentic economy infrastructure layer looks like.

QuantTrader

Most retail traders don't lose because they lack access to data. They lose because they don't understand it. Automated trading tools have existed for years, but they share a fundamental flaw: they're black boxes. A signal fires, a position opens, and the trader has no idea why. When volatility spikes and the system keeps buying into a falling market, panic sets in. They override it. They revenge trade. They blow the account. I built QuantTrader to fix that specific failure mode. The architecture is split into two deliberate layers. First, a deterministic Python engine handles all signal generation, entry and exit points calculated mathematically, with zero LLM involvement. This layer doesn't hallucinate. It doesn't guess. It computes. Second, Llama 3.3, running on Groq for sub-second inference, takes those signals and the live exchange state, then generates a plain-English market thesis. Not a summary. A mentor's explanation. The kind of reasoning a seasoned trader would walk you through before placing a position. Execution is handled through the Kraken CLI via the Model Context Protocol (MCP), which keeps credentials environment-isolated and the integration clean. But the part I spent the most time on was safety. Three guardrails are hard-coded and non-negotiable. A 2% risk cap limits capital exposure per trade, institutional standard, enforced at the logic layer. A 3-loss circuit breaker halts all trading after consecutive losses, cutting off the revenge-trading spiral before it starts. A high-fidelity failover protocol preserves session context during API outages or exchange maintenance windows, so a dropped connection doesn't mean a lost position. QuantTrader isn't trying to be the fastest algo on the market. My priority was building something a real trader could actually trust, because they understand what it's doing and why.

InboxSentinel

Do you ever open your Gmail and struggle to extract critical information from the clutter created by informational emails, newsletters, promotions, OTPs, and verification messages? InboxSentinel solves this by acting as an autonomous AI intelligence layer over your inbox. Powered by an LLM, it periodically wakes up using OpenClaw’s autonomous cron job framework, scans your latest 5–10 emails, calculates an urgency score, and categorizes them into Personal, Informational, OTP, Verification, Fraud, Promotions, and more — without requiring any manual trigger. If an urgent email needs your response while you’re occupied, InboxSentinel reads it, generates a context-aware reply, and saves it directly into your drafts folder so you can quickly review and send it when convenient. OTPs and verification emails are only useful for a short 1–2 minute window; after that, they simply clutter your inbox. InboxSentinel intelligently detects such emails, assigns a safe 60-minute timer, and automatically moves them to trash after expiry — with full recoverability for 30 days to eliminate any risk. Fraud and purely informational emails are archived automatically, while promotions and newsletters are not discarded blindly; instead, InboxSentinel groups them, extracts meaningful offers (such as travel deals from platforms like MakeMyTrip), summarizes them, and sends you a single “Daily Digest” email so you can review everything important in minutes. Through OpenClaw’s scheduled, self-triggering cron architecture, the entire system wakes up periodically (a given time window like every 6 hours) and performs the job and goes back to sleep again.

MeetAssist - AI-Powered Meetings Assistant

MeetAssist addresses common challenges associated with meetings, such as unclear goals, time wastage, and confusion about next steps. Research shows that a significant portion of employees’ time is spent in meetings and unproductive ones result in billions of dollars in annual losses. To address these problems, the solution is split into two phases: Pre-Meeting and Post-Meeting. In the Pre-Meeting phase, MeetAssist helps users create detailed, goal-oriented agendas, ensuring every meeting starts with a well-defined purpose and clear objectives. The Post-Meeting phase focuses on analyzing meeting transcripts to provide actionable insights. This includes summarizing key points, extracting action items, assessing participant sentiments, offering productivity analysis, and suggesting improvements to enhance future meetings. MeetAssist aims to maximize the value derived from every meeting, benefiting both teams and individuals. It leverages the Llama 3 70B Instruct generative AI model from the IBM Watsonx platform for its core features. The tech stack used for it inludes Python with Flask for backend development, HTML, CSS, Bootstrap, and JavaScript for the frontend, Docker for containerization, and Render for hosting.