Llama 4

Top Builders

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

Llama 4

Llama 4 is Meta AI’s newest open-weight model series.
It introduces Mixture-of-Experts (MoE) routing for efficient inference, accepts both text and images natively, and stretches context windows to record-breaking lengths—10 M tokens in the Scout variant. Meta positions Llama 4 as a research-friendly, production-ready alternative to proprietary frontier models, while keeping the code and weights downloadable from its GitHub repos and the official llama.com portal.

General
Release date	5 Apr 2025
Developer	Meta AI
Type	Open-weight multimodal LLM
License	Llama 4 Community License
GitHub	meta-llama/llama-models

Core Features

Mixture-of-Experts architecture – Each query activates a subset of specialised “experts,” yielding higher throughput per FLOP while scaling to trillions of total parameters (TechCrunch).
Native multimodality – Models ingest both text and images without external adapters (The Verge).
Extended context windows – Scout handles up to 10 M tokens; Maverick supports 1 M tokens (llm-stats).
Multilingual training – Optimised across 200+ languages for global deployments (Data Scientist Guide).
Fine-tunable & agent-ready – Models ship with recipes for supervised fine-tuning, LoRA, and RAG inside the Llama Cookbook.

Model Variants

Variant	Active Params	Experts	Total Params	Context Window	Best for
Scout	17 B	16	109 B	10 M tokens	Long-context RAG, document analysis (stats)
Maverick	17 B	128	400 B	1 M tokens	Coding & reasoning tasks, general chat (Oracle Docs)
Behemoth	288 B*	16	~2 T	TBA	High-end STEM, under training (not yet released)

Tools & Resources

Weights & License – Download from the official portal.
Inference & training code – meta-llama/llama-models.
Quick-start recipes – Llama Cookbook.
Stats & benchmarks – Interactive dashboards at llm-stats.com.
Cloud endpoints – Ready-to-use deployments in Oracle OCI Generative AI.

Ecosystem & Integrations

Meta AI assistant now runs Llama 4 across WhatsApp, Messenger, Instagram, and web chat (The Verge).
OCI Generative AI offers managed Scout & Maverick endpoints for enterprise workloads (Oracle Docs).
Community hosting – Providers such as DeepInfra, Groq, and Together price Llama 4 as low as $0.08 / 1 M input tokens (llm-stats).
Research & open-source – Thousands of fine-tuned checkpoints already live on Hugging Face; Meta’s annual LlamaCon (29 Apr 2025) spotlights academic collaborations (TechCrunch).

Llama 4 pushes open-weight LLMs into frontier-model territory—combining trillion-scale capacity with permissive licensing. Start experimenting by cloning the GitHub repo, reading the cookbook, or provisioning a managed endpoint on Oracle OCI.

Edit on GitHub

Meta Llama 4 AI technology Hackathon projects

Discover innovative solutions crafted with Meta Llama 4 AI technology, developed by our community members during our engaging hackathons.

TELMED AI DOCTOR

Telmed AI Doctor is an AI-powered telemedicine platform built to close Africa's healthcare access crisis. Nigeria has only 0.4 doctors per 1,000 people. Millions suffer from treatable conditions simply because qualified medical guidance is out of reach. Telmed deploys AI as the first line of medical response — available 24/7, on any device, at zero cost to the patient. Users interact via text, voice, or image. They describe symptoms naturally or photograph affected areas and receive a structured clinical interpretation, severity classification — Low, Moderate, or Urgent — a personalised medication schedule, and clear escalation triggers indicating when a human doctor is immediately required. The AI engine runs on a multi-model cross-validation architecture. Google Gemini handles primary diagnostic reasoning while DeepSeek independently validates the same input, eliminating hallucination risk on clinical outputs. Google Cloud Vision powers image-based symptom analysis. Google Places connects users to the nearest verified pharmacy to complete the full care pathway. Built on Node.js and Express.js with a vanilla JavaScript frontend optimised for low-bandwidth African mobile networks, every decision was made for real-world deployment conditions. Telmed is not a product. It is infrastructure. The kind Africa should have built a decade ago.

IBM-BOB-Haackathon-submission-BOB_RAG

BM BOB Agentic Framework (Custom) BOB Agentic Framework This project uses the BOB Framework for AI-assisted development. The framework coordinates specialized AI modes to handle implementation, validation, and documentation tasks. Agent Roster 🔀 Orchestrator: Coordinates workflow and manages task delegation 💻 Code Mode: Implements features and writes code ✅ Validator: Verifies implementations and runs tests 📝 Documenter: Generates feature documentation Workflow The framework follows a structured workflow: Orchestrator analyzes requests and creates TODO lists Code mode implements each task Validator verifies implementation immediately Process repeats with retry protocol (max 3 attempts) Final validation and git diff review Optional pull request creation Documentation generation Key Features Context Caching: Reduces redundant file reads across mode switches Incremental Validation: Catches issues early with immediate verification Bounded Execution: Caps retries at 3 attempts with diagnostic analysis Safety Mechanisms: Blocks destructive commands requiring user approval GitHub Integration: Auto-generates PR descriptions from git diffs Jira Integration: Fetches requirements and posts completion summaries Framework Documentation For detailed information about the BOB framework configuration:

PHRAXIS - Voice to Production Quantum Opt System

Every software team knows the pattern: a 10-minute meeting where everyone agrees on a feature, then nothing ships for weeks. PHRAXIS removes that gap entirely. A developer speaks a feature request out loud, and PHRAXIS delivers a production-ready, quantum-optimized GitHub pull request in under 4 minutes. PIPELINE: 1. SPEAK - Developers describe features naturally through browser voice input, with optional text fallback. 2. TRANSCRIBE - IBM Watson Speech-to-Text converts audio into structured text with timestamps and confidence scoring. 3. EXTRACT - IBM Natural Language Understanding identifies developer intent, target modules, parameters, and constraints. Structured output is stored in IBM Cloudant. 4. QUANTUM OPTIMIZE - IBM Quantum QAOA calculates the optimal subset of files to modify before generation begins. Classical evaluation across N candidate files requires 2^N combinations. At enterprise scale (N=50), that exceeds 1 quadrillion possibilities. IBM Quantum evaluates the solution space simultaneously and returns the conflict-free optimal change set, directly guiding IBM Bob’s planning workflow. 5. GENERATE - IBM Bob reads the full repository in Architect mode, receives the quantum-selected files, plans implementation in Plan mode, and writes production-ready code in Code mode that matches existing repository patterns instead of generating generic boilerplate. IBM Granite via watsonx.ai pre-analyzes the codebase, while watsonx Orchestrate coordinates the pipeline. 6. SHIP - Bob Shell executes /review on generated code, and the GitHub API opens a pull request containing the transcript, optimization result, and Bob session IDs. IBM Bob is the execution engine behind PHRAXIS. IBM Quantum determines exactly which files Bob modifies, producing mathematically conflict-free pull requests at enterprise scale.

PR Party

PR Party reimagines the code review experience as a turn-based RPG encounter. When a developer submits a pull request, six specialist AI characters spring into action, each analyzing the diff from their unique domain of expertise. Aegis, the Security Paladin, hunts for vulnerabilities and hardcoded secrets. Schema, the Database Mage, detects N+1 queries and unsafe migrations. Pixel, the UX Bard, flags missing ARIA labels and broken loading states. Atlas, the Architecture Explorer, identifies god functions and tight coupling. Echo, the Test Priest, finds untested code paths and meaningless assertions. Codex, the Docs Scribe, catches outdated references and missing JSDoc. Each character runs as a parallel LLM call against IBM watsonx.ai, analyzing the pull request diff simultaneously. Findings are routed to their specialist island, voiced in the character's personality, and presented as damage dealt to the PR's health bar. The result is a verdict: Approved, Changes Required, or Blocked. The frontend is built with Next.js and features animated floating islands, sprite-based characters, and real-time finding cards that appear progressively as the analysis completes. Developers navigate between the six islands to explore each character's findings.

Agentic Security & Observability Platform

Governance.AI is an agentic security and observability platform designed for autonomous AI systems and multi-agent workflows. As AI agents become more integrated into real-world products, most systems still operate with very little visibility, monitoring, or governance. We wanted to build infrastructure that helps developers understand and control how AI systems behave internally. The platform introduces a centralized governance layer between users and AI agents. Instead of acting like a simple chatbot wrapper, Governance.AI continuously monitors workflows, traces execution paths, analyzes risky behavior, and enforces governance policies before actions are executed. Core capabilities include: Risk Detection & Prompt Analysis Policy Enforcement & Access Control Agent Monitoring & Observability Audit Trails & Explainability Red-Team Testing & Unsafe Prompt Detection Real-Time Governance Workflows The platform is built using a modular FastAPI service architecture connected through a scalable API gateway. We integrated LangSmith for tracing and observability, Auth0 for authentication, Neon PostgreSQL for infrastructure data, and a real enterprise-style dashboard with live workflows, analytics, traces, and governance events. Governance.AI can be integrated using: Python SDK REST APIs Gateway-based integrations Developers can test governance services directly from the dashboard, inspect traces, monitor workflows, and integrate Governance.AI into their own AI systems using SDKs or APIs. One important aspect of this project is that we intentionally avoided building a purely mocked prototype. Our focus was building a realistic developer infrastructure platform that could evolve into a production-grade governance layer for future AI ecosystems. We believe governance, trust, and observability will become foundational infrastructure for autonomous AI systems, similar to how monitoring and security became essential for cloud computing.