solutions.compute_routing

Compute routing
below official list · long-term

One routing layer for mainstream CN and global models. Settlement below official list prices — built for software vendors, companies, and studios on long-term volume. Center-to-center exchange, supply aggregators, or we build your internal API gateway. Volume and terms in writing after inquiry.

Below official listCN + global catalogLong-term routingInternal gateway

One point: routed settlement is below each model’s official public list. Software, companies, and studios can sign long-term volume. If you have no gateway, we build an internal aggregation API under your brand.

price

Core: cheaper than official

Not a discount sheet — contract settlement below vendor list. Larger long-term volume, stabler terms.

  • ▹Same model, routed settlement below official public list — in writing, no “X% off” table on this page
  • ▹Bill by token, call, or hours — fits products, internal systems, and studio workloads
  • ▹New models join the same gateway; you do not reopen every official account to renegotiate
  • ▹No consumer top-up or C-end keys on this site; corporate contracts and invoices

Who it fits

  • ▹Software firms embedding GPT / Claude / CN models in a product, on long-term routing
  • ▹Companies that want one internal model entry for teams and subsidiaries — not a card per person
  • ▹Studios and indie teams that want one internal API across vendors, with statements
  • ▹Compute centers exchanging spare or deficit capacity, or supplying an aggregator

We will not take

  • ▹A consumer mall: buy keys online, wallets, public “X% off” calculators
  • ▹Unauthorized official seats, shared logins, or terms bypass
  • ▹A “cheapest screenshot” with no volume, authorization, or billing entity
  • ▹One-off scalping or personal top-up that cannot invoice
scenarios

Four paths

Exchange, supply, long-term routing for software/companies, or we stand up your internal aggregation platform.

Center-to-center

Compute center ↔ compute center. Idle hours or authorized calls move to the short side — routing plus a statement.

  • ·Auth and channel isolation
  • ·Rate limits / breakers / fallback
  • ·Daily or monthly settlement fields

Supply a platform

Center → aggregator. They get one call entry and usage; they do not wire every channel you run internally.

  • ·Unified endpoint
  • ·Catalog limited to authorized scope
  • ·Settlement cycle in writing

Software / enterprise, long-term

Wire models into a product, internal system, or mid-office. Contract volume below official list; new models can join later.

  • ·In-product calls
  • ·Corporate invoicing
  • ·Stabler at higher volume

Internal API aggregation

A gateway for a company, team, or studio: auth, routing, metering, optional white-label. Staff or customers hit only your API.

  • ·Enterprise / group gateway
  • ·Studio and indie teams
  • ·Branded domain or on-prem
catalog

CN and global catalog

Names we can assess today. New versions and vendors append by contract. Exact SKUs and whether they are live are confirmed in writing after inquiry.

CN · text / multimodal

Qwen3 / Qwen-Max / Qwen-Plus / Qwen-Flash / Qwen-Coder / QwQ / Qwen-VLGLM-4 / GLM-4.5 / GLM-4.6 / GLM-4-Plus / GLM-4-Flash / GLM-4V / ChatGLMKimi / Moonshot K2 / moonshot-v1DeepSeek-V3 / DeepSeek-R1 / DeepSeek-Chat / DeepSeek-CoderDoubao-pro / Doubao-lite / Doubao-1.5-pro / Doubao-1.5-thinking / SeedERNIE 4.0 / 4.5 / Speed / LiteSpark 4.0 / Max / ProHunyuan-Turbo / Hunyuan-Large / Hunyuan-T1MiniMax abab6.5 / abab7 / M2Step-2 / Step-1VSenseNovaBaichuan4 / Baichuan3Yi-Lightning / Yi-LargeInternLM / InternVLSkywork360GPTHuawei Pangu

Global · text / multimodal

OpenAI GPT-4o / GPT-4o-mini / GPT-4.1 / GPT-4.1-mini / GPT-5 / o1 / o3 / o3-mini / o4-miniAnthropic Claude Opus 4 / Sonnet 4 / Haiku 4.5 / 3.7 Sonnet / 3.5 SonnetGoogle Gemini 2.0 Flash / 2.5 Flash / 2.5 Pro / 3 Pro / 1.5xAI Grok-2 / Grok-3 / Grok-4Meta Llama 3.1 / 3.3 / 4Mistral Large / Mixtral / CodestralCohere Command R+Amazon Nova Pro / Lite / MicroMicrosoft Phi-4Perplexity SonarNVIDIA Nemotron

Image

GPT Image / DALL·E 3Flux 1.1 Pro / Flux 2MidjourneyStable Diffusion 3 / SDXLImagen 3 / 4Ideogram 3Recraft V3KolorsJimeng / SeedreamQwen-ImageGemini image / Nano Banana

Video

Sora 2Kling 1.6 / 2.0 / 2.1 / 2.6 / 3Runway Gen-3 / Gen-4Luma RayPikaViduHailuo / MiniMax videoSeedanceWan 2.1 / 2.2 / 2.6PixVerseGoogle Veo 2 / 3Grok Imagine video

Speech / music

OpenAI Whisper / GPT-4o AudioElevenLabsMiniMax SpeechSuno / UdioCosyVoiceFish AudioAzure TTS
partners

Long-term routing

Not a one-shot dump. Products, IT, and studios can pull monthly or yearly volume into one layer.

Software and SaaS

Embed many models in support, writing, code, review, agents. One key policy. Iterate without reopening every official account.

Companies and groups

One internal entry, with permission, audit, and statements on your side. On-prem optional. Knowledge bases, office assistants, process calls.

Teams and studios

Design, content, overseas, and R&D shops collapse CN and global models into an internal platform. Bill by project or month.

grades

What the internal gateway covers

Companies, studios, and software teams can expose “one API inward”. Capabilities open per contract.

01

One call surface

One auth and path in front of CN text, global text, multimodal, speech, image, and video. Callers do not see how many vendors sit behind.

02

Routing and fallback

Switch by availability, latency, or your rules. Failover is possible; no SLA numbers on this page.

03

Metering and statements

Bill by calls, tokens, or agreed units. Split by product line, department, or project.

04

White-label and on-prem

Your domain and brand, inward or outward — or deploy on your side. Source and ops boundaries go in the contract.

audience

How you connect

You already have keys / accounts

  • ▹Vault or stay on your side
  • ▹We route, meter, and attach below-list channels
  • ▹You stay inside vendor terms

You want our routing channel

  • ▹Authorized CN and global models
  • ▹Settlement below official list, in writing
  • ▹Fits software, companies, studios

You want an internal gateway

  • ▹Gateway under your brand
  • ▹Staff/customers hit only your API
  • ▹On-prem optional; hook your software
process

How it starts

01 · Intent

Software, company, studio, or compute center; long-term routing vs internal platform; CN/global and daily volume. Form: “Compute routing”.

02 · Scope

Fit, path, catalog, and how far below official list — in writing within 24 hours. No public “X% off”.

03 · POC

One path: auth, call, meter. POC can be billed alone.

04 · Attach or build

Add models, statements, white-label or on-prem by milestone. Software integration can stack as a separate contract.

05 · Long-term settlement

Metering rules, statement cycle, responsibility. Monthly or yearly. New models append.

deliverables

Typical deliverables

  • Scope and authorization checklist
  • Call notes (path, auth, error codes)
  • Metering and statement field notes
  • Optional: internal API aggregation (white-label or on-prem)
  • Long-term routing terms vs official list (written)
  • Ops and upgrade boundary (contract)
related

Related

Custom / overseas →Software development →Dedicated line (connectivity) →Strategic partnerships →

All articles →

API aggregation sites and custom channelsWhite-label deliveryGateway vs iPaaS

This page describes model routing and gateway work. “Below official list” means contract settlement, not a resale license from any vendor and not a public price list. We only attach authorized accounts and capacity. No self-serve keys or top-up on this site. The catalog tracks the market; the written inquiry list wins.

Yes. Routed settlement is below the model’s official public list — that is the point of this page. The gap is confirmed in writing by volume and SKU. We do not publish an “X% off” table.
// compute.route()

Role and volume first
then below-list routing

Software, company, studio, or compute center — long-term routing or an internal gateway, plus daily volume. Path and terms in 24 hours.

// CONTACT
Need help? Talk to an advisor
WeCom live chat · WeChat QR
Site emergencyMalware / red screen ▸
在线客服Live reply ▸
中EN