Skip to main content
Runcrate logo

Llama 3.3 70B API pricing on Runcrate

Every Runcrate Llama 3.3 70B rate we track, compared against 14 providers serving the same model.

Input from /1M
$0.105
Output /1M
$0.336
Input price rank
#2 of 14
Last updated
November 21, 2026

The cheapest Llama 3.3 70B input price we currently track is $0.100/1M on Deep Infra.

The cheapest Llama 3.3 70B output price is $0.320/1M on Deep Infra.

Runcrate Llama 3.3 70B rates

Prices per 1M tokens. Batch and cached-input rates are shown where the provider publishes them.

Runcrate Llama 3.3 70B pricing by mode
ModeInputOutput
Standard$0.105/1M$0.336/1M

About Llama 3.3 70B

Creator
Meta
Context
128K
Modalities
text
Knowledge cutoff
Mar 2024
Tool calling
Yes
Open weights
Yes
Full Llama 3.3 70B details and every provider →Compare API pricing for every Llama model →

Llama 3.3 70B on other providers

About Runcrate

Runcrate runs a token-billed inference API over a catalog of open-source text, image, video, audio and embedding models alongside on-demand GPU instances and reserved clusters, billed from a single credit balance. It is operated by Aeonmind, LLC.

Frequently Asked Questions

How much does Llama 3.3 70B cost on Runcrate?

Runcrate serves Llama 3.3 70B from $0.105 per 1M input tokens, across 1 tracked pricing mode. Prices are collected daily — see the table above for current input, output, and cached-input rates.

Is Runcrate the cheapest way to run Llama 3.3 70B?

Runcrate ranks #2 of 14 providers we track serving Llama 3.3 70B on input price. The cheapest input price right now is on Deep Infra. The lowest output price is $0.320/1M on Deep Infra. Rates change often, and throughput, latency and context limits differ between providers — check the comparison above before committing.

Does Runcrate offer batch or cached pricing for Llama 3.3 70B?

We track 1 Llama 3.3 70B offering from Runcrate. Batch and cached-input rates appear in the table above when the provider publishes them; blank cells mean we have no current figure for that mode.