Llama 3.3 70B is tracked as self-hosted / varies. This catalog does not publish a universal first-party hosted API price for the model; the official channel is labeled self-hosted or pending. The OpenRouter channel remains separate so the two routes are not mistaken for one price.
What is Llama 3.3 70B in this model directory?
Llama 3.3 70B is a Meta model tracked with self-hosted / varies. This directory page brings together its current official and OpenRouter routes, context window, lifecycle notes, verification dates, example cost, and same-provider alternatives in one source-linked record.
The cost position is a directory aid, not a capability ranking. A lower sample price does not mean two models are interchangeable.
What are the Llama 3.3 70B input and output prices?
| Channel | Input / 1M | Cached input / 1M | Output / 1M | Verified | Source |
|---|---|---|---|---|---|
| Official direct | — | — | — | — | Provider source ↗ |
| OpenRouter | $0.10 | — | $0.32 | 2026-09-04 | — |
Rates are normalized to USD per 1M tokens. This page does not replace a missing official price with an OpenRouter price.
What does Llama 3.3 70B cost for a typical workload?
Using 1M input tokens and 250K output tokens, the estimated cost is — through the official route and $0.18 through OpenRouter. This example is calculated from the listed token rates and does not include taxes, tools, storage, retries, or optional OpenRouter credit purchase fees.
example = (1M × input rate) + (0.25M × output rate)
monthly estimate = example × your monthly runs
Open the full workload calculator to change the token mix, cached-input percentage, monthly runs, and selected channel.
What are the context and lifecycle details?
Capacity is not a flat fee; billed tokens depend on what a request actually processes.
Open-weight model. Meta does not publish a single first-party per-token API price; hosting costs vary by provider and hardware.
0 official and 4 OpenRouter observations are stored for this model.
Which Meta models are closest for comparison?
These same-provider models are selected by similar example workload cost where comparable prices exist. Use them as navigation shortcuts, not as a claim of equivalent capability.
| Model | Official in / out | OpenRouter in / out | Context |
|---|---|---|---|
| Llama 4 Maverick | No first-party API | In $0.1875Out $0.6525 | — |
| Llama 4 Scout | No first-party API | In $0.10Out $0.30 | — |
How is this Llama 3.3 70B price page maintained?
The official route uses the provider-owned source shown above. The OpenRouter route uses an explicit model-ID mapping and the exact OpenRouter record shown above. Source failures preserve the last verified record; a price-history event is added only when a verified input, cached-input, or output rate changes.
For methodology, source failure handling, and channel definitions, see the pricing methodology and price history.
Llama 3.3 70B pricing questions
How much does Llama 3.3 70B cost per 1M tokens?
This catalog does not publish a universal first-party hosted API price for the model; the official channel is labeled self-hosted or pending. The mapped OpenRouter route is $0.10 input and $0.32 output per 1M tokens.
What is the Llama 3.3 70B example workload cost?
For 1M input tokens and 250K output tokens, the estimated cost is — through the official channel and $0.18 through OpenRouter, before any optional OpenRouter credit purchase fee.
Does Llama 3.3 70B have an OpenRouter route?
Yes. The route is stored separately under its OpenRouter model ID and has its own source and verification date.
What should I check before using Llama 3.3 70B in production?
Confirm the linked provider terms, context and long-context rules, cache behavior, rate limits, latency, tool support, data policies, and actual token usage for your workload. This page is a pricing reference, not a capability or procurement recommendation.