AREX Turbo is a chat model created by BAAI. It is an open-source, text-in, text-out model designed as a compact research agent for long-horizon tasks involving search, evidence verification, and recursive self-improvement.
Input
Output
Context
262K
Max Output
-
Parameters
4.5B
Input Modalities
Output Modalities
None the catalog can show from stated facts or measured scores.
Download 9.08 GB · Runs on vLLM v0.32.0 · 2 GGUF builds
| Precision | Weights | One request at 33K tokens | Smallest GPU that fits |
|---|---|---|---|
| FP16 / BF16 | 8.46 GB | 12.1 GB | 1x NVIDIA T4 (16 GB) |
| INT8 | 4.23 GB | 7.69 GB | 1x NVIDIA T4 (16 GB) |
| INT4 | 2.11 GB | 5.47 GB | 1x NVIDIA T4 (16 GB) |
Computed by Inferbase from the 4,539,265,536 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 10, 2026.
Answered from the facts on this page, with their sources.
AREX Turbo has a context window of 262,144 tokens (source: Creator or provider API). A provider may serve less; Inferbase routes on what each provider serves.
Yes. AREX Turbo is an open-weight model under the Apache 2.0 license. You can download and self-host it within the license terms.
AREX Turbo has 4,539,265,536 parameters (4.5B), as stated in the creator's Hugging Face repository.
Strengths: Open weights under a permissive license (apache-2.0); 262K token context window. Each is computed from the model's stated facts and benchmark results.
AREX Turbo accepts Text and produces Text. The Capabilities card lists the API features it supports, such as function calling and structured output.
No benchmark source has published a result for AREX Turbo yet, so it is shown as unmeasured, not estimated.
Not on Inferbase Managed today. Add your own provider key or endpoint that serves AREX Turbo, and Inferbase runs it on your account with the same routing and logs.
The weights take 8.5 GB at FP16 / BF16, 4.2 GB at INT8, 2.1 GB at INT4, computed from the 4,539,265,536 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.
The model card of AREX Turbo lists one language: English. That is what BAAI states; the catalog does not test language quality.
Start on the free plan with 5,000 requests a month, no card required.