Meta
Llama 4 & Muse Spark — the world's most widely deployed open-weight models
Meta AI is the creator of the Llama model family, the most widely used open-weight LLMs in the world. Llama models are available via Meta's own API and through dozens of third-party inference providers. The Llama 4 series includes Behemoth (2T params), Scout, and Maverick, with 1M-token context windows. Meta also offers Muse Spark, a proprietary multimodal model. Because Llama weights are open, teams can self-host on GPU cloud for dramatically lower per-token costs at scale.
The only frontier-class model family available as open weights — enabling self-hosted inference on GPU cloud at a fraction of API pricing for high-volume workloads.
- Open-weight models — self-host on any GPU cloud for lowest per-token cost at scale
- Llama 4 Behemoth: 2T parameter frontier model with 1M context window
- Widest third-party hosting ecosystem — available on AWS, Azure, GCP, Together AI, Groq, and 20+ others
- Llama 3.2 1B/3B models run on-device (mobile, edge)
- No vendor lock-in — switch inference providers without changing model weights
- Self-hosting requires GPU infrastructure expertise
- Meta's own API has limited availability vs third-party hosts
- Llama 4 Behemoth pricing not yet publicly listed
- Smaller proprietary model lineup vs OpenAI/Anthropic
All Models
Meta — Frequently Asked Questions
Community Reviews
All LLM Providers
Full comparison tableMeta LLM pricing overview
Meta is a proprietary AI lab that develops and hosts its own large language models. Meta AI is the creator of the Llama model family, the most widely used open-weight LLMs in the world. Llama models are available via Meta's own API and through dozens of third-party inference providers. The Llama 4 series includes Behemoth (2T params), Scout, and Maverick, with 1M-token context windows. Meta also offers Muse Spark, a proprietary multimodal model. Because Llama weights are open, teams can self-host on GPU cloud for dramatically lower per-token costs at scale. Common use cases include Self-hosted inference, Cost-optimised at scale, Edge/on-device, Chat. Headquartered in Menlo Park, CA, founded 2023. All models are accessible via a REST API compatible with standard OpenAI-style request formats, enabling drop-in integration with most LLM frameworks and orchestration tools.
Meta vs other LLM providers
Meta competes with OpenAI, Anthropic, Google, Groq, Together AI, Mistral, Cohere, and other inference API providers across dimensions of price, throughput, latency, context window, and model intelligence. The only frontier-class model family available as open weights — enabling self-hosted inference on GPU cloud at a fraction of API pricing for high-volume workloads. Use the LLM provider comparison tool to see Meta token pricing, latency, and throughput side-by-side with any other provider. The full LLM pricing table shows all providers ranked by input token cost, output token cost, and throughput in a single sortable view.
Why choose Meta?
Meta's key strengths are: Open-weight models — self-host on any GPU cloud for lowest per-token cost at scale; Llama 4 Behemoth: 2T parameter frontier model with 1M context window; Widest third-party hosting ecosystem — available on AWS, Azure, GCP, Together AI, Groq, and 20+ others. Limitations to consider: Self-hosting requires GPU infrastructure expertise; Meta's own API has limited availability vs third-party hosts. For teams running high-volume inference workloads, prompt caching and batch API endpoints can reduce effective input token costs by 50–90% — check the context window cost guide for a full breakdown of caching economics.
Understanding Meta token pricing
Meta charges separately for input (prompt) and output (completion) tokens, priced per 1M tokens in USD. Output tokens are typically 3–5× more expensive than input tokens due to the compute cost of autoregressive generation. Prices shown are sourced from Meta's public pricing page and updated daily. Need help estimating your spend? Read the LLM API cost calculator guide — it covers tokens, context windows, prompt caching, and batch discounts with worked examples for RAG, chat history, and document processing workloads.
Meta context window and model capabilities
Context window size directly affects both capability and cost — every token in the context window is charged as an input token. For RAG and document processing workloads, longer context windows enable richer retrieval but increase per-call costs proportionally. Prompt caching — where supported — stores the KV state of repeated prefixes and charges 75–90% less for cache hits, making it the most impactful cost optimization for applications with consistent system prompts or retrieved documents. See the LLM context window cost guide for a full analysis of how context length affects your API bill.
Self-hosted vs managed inference: when Meta makes sense
Managed inference APIs like Meta eliminate infrastructure overhead — no GPU provisioning, driver management, or model serving stack to maintain. The trade-off is cost at scale: a single H100 at ~$2.50/hr can serve ~500K tokens/min of Llama 3.3 70B, which at Meta API rates would cost significantly more per token. The break-even point depends on your request volume, latency requirements, and engineering capacity. For teams processing fewer than ~10M tokens/day, managed APIs are almost always cheaper when total cost of ownership is considered. Above that threshold, self-hosted inference on rented GPU compute typically wins on unit economics. Read the cheapest GPU cloud guide for a full break-even analysis. Historical Meta token price data is available in the LLM price history charts.