Jiuzhang Alaya Cloud
Alaya Token — three-tier personal subscription aggregating Kimi K3, GLM-5.2, and DeepSeek-V4 Flash
Why we recommend Jiuzhang Alaya Cloud
Why we recommend Jiuzhang Alaya Cloud
Jiuzhang Alaya Cloud (alayanew.com) is the public model-service brand of DataCanvas (九章云极), a Chinese data-science infrastructure vendor. The core product, Alaya Token (TokenPlan), aggregates three flagship domestic models — Kimi K3, GLM-5.2, and DeepSeek-V4 Flash — under a single subscription, billed per token. The product positioning is "aggregate without lock-in".
Multi-domain note. www.alayanew.com is the brand portal (model announcements, Alaya NeW Cloud positioning). codingplan.alayanew.com is the Alaya Token / TokenPlan subscription page — the personal Lite / Plus / Max tiers live here, and the "Subscribe now" CTA on the detail page points at this domain. docs.alayanew.com is the developer docs hub (API reference, IDE integration, billing reference). Console login routes through sso.alayanew.com; OAuth callback lands on
www.alayanew.com/callback. A single SSO account works across all subdomains.
Three personal subscription tiers
TokenPlan ships three flat-monthly tiers for individual developers — all share the same model pool and token-based billing, differing only in target workload:
| Tier | Monthly | Target workload |
|---|---|---|
| Lite (basic) | ¥199 / mo | Individuals on everyday lightweight use; writing polish, study Q&A, document org; personal projects, small-repo light iteration |
| Plus (advanced) | ¥399 / mo | Power users, creators, developers on daily high-frequency use; batch copywriting, long-document processing, data analysis; mid-size project dev |
| Max (power) | ¥699 / mo | Team collaboration and enterprise-scale; knowledge-base Q&A, ultra-long document processing, Agent workflows; large repos, enterprise R&D |
All three tiers include full access to Kimi K3, GLM-5.2, and DeepSeek-V4 Flash with real-time token deduction; new models from upstream typically auto-join the pool without a fresh subscription.
Note: as of the 2026-08-16 review, the three personal tiers (Lite / Plus / Max) display "售罄" (sold out) on the subscription page and the new-purchase entry is temporarily closed; watch for reopening or contact sales. The enterprise tiers and custom tier remain open.
Flagship model lineup (three flagship under one subscription)
| Model | Positioning |
|---|---|
| Kimi K3 (Moonshot AI) | 1M context, 2.8T-MoE architecture, native image + video, strong on long-horizon agents |
| GLM-5.2 (Zhipu AI) | Open-source flagship tuned for long-task stability |
| DeepSeek-V4 Flash | 1M context, next-gen reasoning model; stronger planning, math, and code |
"Aggregate without lock-in" is the product's recurring message — one subscription lets you switch the lead model by scenario rather than paying and integrating per model.
100% OpenAI / Anthropic dual-protocol compatible
Alaya Token exposes both upstream-standard bases — OpenAI-compatible and Anthropic-compatible — so existing code ports over with a single baseURL plus a Key swap (the OpenAI-compatible endpoint is https://token.alayanew.com/v1, per the docs example /v1/chat/completions):
| Tool | Integration point |
|---|---|
| OpenClaw (Jiuzhang's in-house IDE) | Settings > API Key |
| Claude Code (Anthropic) | ANTHROPIC_BASE_URL env var |
| Codex (OpenAI) | ~/.codex/config.toml > base_url |
| OpenCode (open-source terminal) | opencode.json > base_url |
No need to write a new integration for each tool — the vendor calls this "one-line configuration" out to developers.
Domestic direct-connect, low latency, no VPN needed
Service runs entirely on domestic infrastructure and is reachable over default DNS from inside China with no proxy required. The recurring failure modes for overseas services (request timeouts, 401s, "region not available") essentially don't apply to Alaya — and that's the implicit prerequisite for the Max tier's "team collaboration and enterprise-scale workload" framing.
Enterprise tiers and custom
TokenPlan adds three enterprise tiers: Light ¥2,999 / month (up to ~2.98B tokens / month), Standard ¥3,999 / month (up to ~3.98B), and Pro ¥9,999 / month (up to ~9.95B). All three ship with first-month bonus allocations (15% / 20% / 50%) and ongoing top-ups (5% / 10% / 20% from month two) — designed to nudge enterprise callers to consolidate model spend on TokenPlan. Above that sits Custom — model mix, seat count, and SLA all configured per engagement, pricing by quote, contact via the official site or sales.
If your project needs a single subscription covering all three flagship Chinese models + one-line IDE integration + reliable domestic connectivity + a personal-to-enterprise tier ladder + freedom to switch lead models, Alaya Token is a strong 2026 placement on the /models shelf.