Emu 3 Gen is a multimodal model from BAAI that takes text, image, and audio inputs and produces text, image, and audio outputs. It is an open-source, autoregressive transformer designed for tasks like generating images and videos from text, understanding visual content, and extending video sequences.
Input
Output
Context
9K
Max Output
-
Parameters
8.5B
Input Modalities
Output Modalities
None the catalog can show from stated facts or measured scores.
Download 34.0 GB
| Precision | Weights | One request at 9K tokens | Smallest GPU that fits |
|---|---|---|---|
| FP16 / BF16 | 15.8 GB | 17.6 GB | 1x NVIDIA GeForce RTX 3090 (24 GB) |
| INT8 | 7.91 GB | 9.31 GB | 1x NVIDIA T4 (16 GB) |
| INT4 | 3.95 GB | 5.16 GB | 1x NVIDIA T4 (16 GB) |
Computed by Inferbase from the 8,492,011,520 parameters in the creator's Hugging Face repository; one request adds its KV cache, activations and framework overhead, with the layer layout estimated from the parameter count.
Public resources
Specification sources
As of Nov 11, 2026.
Answered from the facts on this page, with their sources.
Emu 3 Gen has a context window of 9,216 tokens (source: Creator or provider API). A provider may serve less; Inferbase routes on what each provider serves.
Yes. Emu 3 Gen is an open-weight model under the Apache 2.0 license. You can download and self-host it within the license terms.
Emu 3 Gen has 8,492,011,520 parameters (8.5B), as stated in the creator's Hugging Face repository.
Strengths: Open weights under a permissive license (apache-2.0). Each is computed from the model's stated facts and benchmark results.
Emu 3 Gen accepts Text, Image and Audio and produces Text, Image and Audio. The Capabilities card lists the API features it supports, such as function calling and structured output.
No benchmark source has published a result for Emu 3 Gen yet, so it is shown as unmeasured, not estimated.
Not on Inferbase Managed today. Add your own provider key or endpoint that serves Emu 3 Gen, and Inferbase runs it on your account with the same routing and logs.
The weights take 15.8 GB at FP16 / BF16, 7.9 GB at INT8, 4.0 GB at INT4, computed from the 8,492,011,520 parameters its creator publishes. Each request adds its KV cache and the framework's overhead. The hardware section shows one request and the smallest GPU that fits; the GPU planner sizes it for your context length and traffic.
Start on the free plan with 5,000 requests a month, no card required.