Skip to main content

​
Short Answer

On APIYI, “Enterprise group” is not a single universal group — each model has its own dedicated enterprise fallback channel. They typically cost a bit more than the default group, but they keep producing images even when the default group is saturated — making them a fallback option for high-success-rate workloads.
This article only lists facts that can be verified from official docs or live status records. SLA, concurrency limits, and other unpublished data should be requested via customer support.

​
Enterprise Groups Are Per-Model

Unlike the universal groups ClaudeCode, Sora2Official, etc., enterprise groups are named per model:
Important clarification: Enterprise groups are NOT synonymous with Vertex fallback. image2Enterprise is an OpenAI upstream fallback — unrelated to Google Vertex; NanoBananaReverse is the channel that actually goes through Vertex (via reverse engineering); NanoBananaEnterprise is a high-availability fallback whose channel composition is not explicitly documented.

​
When the Default Group Saturated, Enterprise Groups Still Worked

Real-world case for gpt-image-2:
  • 2026-05-13 14:29–14:31 call logs: 9 non-streaming calls all routed to the image2Enterprise group, with first-byte latency between 35-293 seconds at 1.2x multiplier — still produced images successfully (see Live Update · 2026-05).
  • When the gpt-image-2 default group was saturated, “some default-group requests had been routed via the fallback path into the enterprise group” — confirming the design intent that enterprise groups serve as a fallback (same live update).
NanoBananaReverse’s design intent (see Live Update · 2026-06):
To relieve the impact on official-direct Nano Banana Pro / 2 during peak hours, we added a reverse-engineered Vertex group NanoBananaReverse at 80% of the default group’s price (0.8x multiplier), per-call billing regardless of resolution.
No “enterprise group = always stable” guarantee. Official docs explicitly state: the gpt-image-2 enterprise group (official OpenAI API) also experienced “The server had an error while processing your request.” errors on 2026-05-07, root-caused to an OpenAI upstream failure (see Live Update · 2026-05). Enterprise groups mitigate default-group congestion but do not guarantee upstream stability forever.

​
How to Switch to an Enterprise Group?

Enterprise groups are configured at the token level:
  1. Open https://api.apiyi.com/token
  2. Find the target token → click the “Manage (wrench icon)” in the action column → “Edit Token”
  3. Under “Select Group”, pick the corresponding enterprise group (e.g. image2Enterprise, NanoBananaEnterprise, NanoBananaReverse)
  4. Save
No code changes needed. Detailed walkthrough in Live Update · 2026-04 · image2-enterprise (gpt-image-2 example; the path is the same for other models).
NanoBananaReverse scope: Only supports gemini-3-pro-image (Nano Banana Pro) and gemini-3.1-flash-image (Nano Banana 2), per-call billing only. If your Nano Banana Pro token is on usage-based billing, this group doesn’t apply.

​
FAQ

Not necessarily — it depends on which model’s enterprise group:
  • image2Enterprise (gpt-image-2 specific) — NOT Vertex, it’s an OpenAI upstream enterprise fallback
  • NanoBananaEnterprise (Nano Banana specific) — docs do not explicitly state whether it uses Vertex
  • NanoBananaReverse (Nano Banana Pro / 2) — definitely uses reverse-engineered Vertex
Don’t equate “enterprise group” with “Vertex fallback”.
Per available docs / live updates:
  • gpt-image-2: image2Enterprise (1.2x)
  • Nano Banana Pro / 2: NanoBananaEnterprise (1.4x), NanoBananaReverse (0.8x)
  • Nano Banana OSS (URL output): NB-OSS (1x, but it’s a beta — you need to contact customer support to enable the group visibility; see Nano Banana OSS Group)
The exact groups available follow what your dashboard shows — the platform may add or rename groups over time.
Typical scenarios from recorded incidents:
  • Default group is persistently saturated / queued / timing out (e.g. gpt-image-2 around 2026-05)
  • Your business is success-rate sensitive and you don’t want default-group failures to drag you down
  • Nano Banana Pro / 2 experience is poor at peak hours — consider NanoBananaReverse (per-call, 0.8x, cheaper)
Conversely, if the default group is well-resourced, don’t bother switching to the enterprise group: on 2026-05-23 after the gpt-image-2 default group was replenished, “call directly — no need to specifically switch to image2Enterprise” (see Live Update · 2026-05).
Not necessarily. Multipliers from recorded data go both ways:
  • image2Enterprise: 1.2x (≈ 20% more than default)
  • NanoBananaEnterprise: 1.4x (≈ 40% more than default)
  • NanoBananaReverse: 0.8x (≈ 20% less than default)
So “enterprise” does not mean “more expensive” — the multiplier depends on the specific model and group.
Yes. A token supports 1 default group + up to 2 fallback groups (see Tokens and Groups). When the primary group is congested, calls automatically switch to a backup channel. This is the recommended pattern for “high success rate” workloads.

Tokens and Groups

Token roles, creation/editing, and rules for default vs. fallback groups.

Nano Banana Developer Guide

Complete Nano Banana documentation, including AIStudio + Vertex dual-channel redundancy.

Live Updates Archive

Daily platform status, channel switches, and incident recovery timeline.

Nano Banana Pricing Overview

Pricing comparison for Nano Banana Pro / 2 / 2 Lite / first generation.
Need more? Information not covered in this article — enterprise-group SLAs, concurrency limits, enterprise-group rollout plans for new models — please contact us via WeCom customer service or email [email protected].