Skip to main content
​
Important · New release (tentative)
The new QWave API portal, catalog and billing rules are liveThe portal, model catalog and documentation have been rebuilt around one idea: one key, every price tier in the open. API paths and API keys are unchanged, so no code changes are needed.What is new
  • Portal and console: home, the model catalog (/explore), pricing (/pricing), rankings and the console are all redesigned. English lives at /, Chinese at /zh-Hans-CN. Old paths under /v2 and /zh redirect automatically.
  • Model catalog: 92 media model groups (34 image, 52 video, 6 audio) plus every text model. Each model’s price tiers, supported actions and parameters are public on its catalog page, comparable by entry price and filterable by vendor and task.
  • One task endpoint: image, video and audio all go through POST /v1/tasks, with action selecting generate, edit, extend and so on, and ref_task_id chaining a follow-up onto an earlier task. Text models keep the OpenAI-compatible /v1/chat/completions.
  • Billing rules: everything is priced in USD. A task freezes the quoted amount on submission; upstream errors, timeouts and content refusals release it automatically. Models settled on token usage (Nano Banana official, GPT Image official, Gemini Omni) settle on actual usage after completion and refund the difference. Every freeze and release is visible in the console’s call log.
  • Documentation: every media model page was rewritten with the same structure (Quick start / Request parameters / Billing / Limits). Model updates and Price updates record every launch, delisting and price change.
Existing API keys, request paths and response shapes are unchanged. Prices on the catalog pages and in Price updates are authoritative; if a locally cached price list disagrees, the catalog wins.
​
Upstream shutdown notice · Gemini Omni Flash preview
gemini-omni-flash-preview will stop on 2026-09-30 with the upstreamGoogle made gemini-omni-1.1-flash generally available on 2026-08-27 and turns the preview endpoint off on September 30, 2026. gemini-omni-flash-preview on QWave API is delisted the same day; tasks submitted afterwards are rejected and not charged.
  • Migration: change model to gemini-omni-1.1-flash; the request shape is the same. Note that the GA model defaults to 720p and adds a 360p draft tier and 4k, each priced differently — see the billing table on the model page.
  • Docs: Gemini Omni Flash
​
Upstream shutdown notice · Sora 2
sora-2 / sora-2-pro will stop on 2026-09-24 with the upstreamOpenAI has announced that the Sora 2 models and the Videos API shut down on September 24, 2026 (announced 2026-03-24). sora-2 and sora-2-pro on this platform are delisted the same day and stop accepting new tasks.
Tasks submitted to sora-2 / sora-2-pro after 2026-09-24 are rejected outright and incur no charge. Tasks submitted and completed before then are unaffected; save your output links in time.
  • Alternatives: for realistic video with a native audio track, consider veo3.1-fast / veo3.1-quality, kling-v3 or seedance-2.0. Each model page documents its request shape; text-to-video and image-to-video actions are available on all of them.
  • Docs: Sora 2 Series
​
Upstream shutdown notice · Speech to text
whisper-1 / gpt-4o-transcribe / gpt-4o-mini-transcribe will stop on 2027-02-26 with the upstreamOpenAI has announced that these three transcription models shut down on February 26, 2027 (announced 2026-08-26) and recommends migrating to gpt-transcribe / gpt-live-transcribe. This platform delists the three models on the upstream shutdown day and will announce the replacement models here ahead of time once they are wired in.
  • Scope: the three models on POST /v1/audio/transcriptions. Speech synthesis via gpt-4o-mini-tts is not affected.
  • Advice: keep model configurable in new transcription integrations so the switch is a one-line change.
  • Docs: Speech to Text
​
Important · Access change
Notice: Web access from mainland China will be closedDear users,Thank you for your continued trust and support. After careful evaluation, due to factors beyond our control, we must make the following adjustment. Please be aware in advance and prepare accordingly.
Starting 2026-07-10, this website will close web page access from mainland China. After that date, qingbo.dev web pages will not open normally within mainland China.
Please note: qingbo.dev will continue to work normally for API calls within mainland China — only the web pages become inaccessible.
In other words, from 2026-07-10, qingbo.dev behaves as follows within mainland China:So users who integrate via the API need not worry about service interruption — only the web frontend is no longer reachable from mainland China.To ensure long-term, stable use of our service, we strongly recommend:
  • Make API requests from overseas or Hong Kong (China) servers where possible;
  • Avoid making calls directly from mainland China servers, for a more stable and durable connection;
  • If you already have workloads deployed in mainland China, migrate or add overseas / Hong Kong nodes early for a smooth transition.
We apologize for any inconvenience and sincerely thank you for your understanding and support. We will keep improving the service for the long run.
​
Service notice · HD tiers
gpt-image-2 2K / 4K temporarily unavailableDue to a temporary technical issue, the 2K / 4K high-resolution tiers of gpt-image-2 are currently unavailable. 1K is unaffected.
If you need 2K / 4K output, please use gpt-image-2 instead — a direct official-channel relay that supports 2K / 4K with higher stability; it is also recommended whenever you need maximum output reliability.
We are working to restore low-cost 2K / 4K on gpt-image-2 and will post an update here as soon as it is back. Thank you for your understanding.