# Zhipu Z Managed Agents

Official pricing: https://docs.bigmodel.cn/cn/managed-agents/overview.md  
Category: agent-sandbox · Isolation: not published

## Pricing regimes (raw)

- **Managed Agent sandbox, temporary free runtime fee** (resource) [alt, beta]

## Features

Yes: snapshots, pause/resume, persistent disk, whole disk survives a stop, volumes, idle auto-stop, ≥24 h sessions, custom image (Docker or snapshot), start from your own snapshot, root, browser, egress allowlist, open internet, Python, Node.js, Go, Rust, Java, credential injection, wake on request, scheduled runs (cron), MCP server, usage per sandbox (API), automatic snapshots, agent harness API, hosted agent API (their own agent), secret proxy, secret proxy for any API, shared volumes

No: memory snapshots, fork/clone, your own Docker/OCI image, Docker inside, nested virtualization, desktop GUI, computer-use API, browser + desktop control, anti-bot stealth, CAPTCHA solving, residential IPs, preinstalled agents, SSH, public IPv4, HTTPS preview URLs, static egress IP, self-hosting, BYOC, open source, SSO, GPU, arm64, Windows, macOS, HTTP method/path egress rules, live resize, live fork (no pause), memory fork, EU data residency, spend limits, snapshots on demand, inbound access rules, extra volumes

Unknown: everything else. Evidence (source + quote) per feature: https://battleships.dev/data/providers/zhipu.json → feature_evidence

## Company

- Ownership: public
- Stage: public [source](https://en.wikipedia.org/wiki/Z.ai)
- Accelerators: none found
- Total raised: $750M (confirmed: Sum of two confirmed pre-IPO rounds reported by Bloomberg/CNBC/SCMP: ~2.5 billion yuan (~US$350M) in Oct 2023 and US$400M in May 2024. Earlier/smaller rounds and exact IPO net proceeds are not captured, so this is a floor, not a full total.) [source](https://en.wikipedia.org/wiki/Z.ai)
- Last round: IPO, 2026-01 (estimated) [source](https://en.wikipedia.org/wiki/Z.ai)
- Investors: Alibaba Group, Tencent, Ant Group, Meituan, Xiaomi, HongShan, Prosperity7 Ventures
- Revenue: unknown

Zhipu AI (rebranded internationally as Z.ai in July 2025; Chinese legal name changed to Knowledge Atlas Technology in 2025) is a Tsinghua University spin-out founded in 2019, one of China's 'AI Tiger' LLM companies, and completed the first IPO among major Chinese LLM firms on the Hong Kong Stock Exchange (ticker 2513) on 8 January 2026, with market cap ~US$62B as of August 2026. It was added to the US Commerce Department's Entity List in January 2025. This card's product is Zhipu's 'Managed Agent' sandbox offering documented at docs.bigmodel.cn, a small part of the broader public company's business; revenue figures here (unknown) would refer to the whole company, not this specific product line.

## Caveats

- Fixed CPU/RAM size, 50-concurrent access and 8-hour runtime eligibility unknown; no 4/8 comparison.
- Runtime promo end date unknown; model-token bill is mandatory but depends on selected model/tokens.
- Outbound network limited to HTTP/HTTPS ports 80/443; allowlists available.
- Unknowns remain null, not free; generic infrastructure modes are opt-in alt.
- Worked example assumes resources are released between active windows unless a retained-instance scenario is explicitly discussed.
- VAT basis: Unknown: reviewed pricing source does not explicitly establish VAT/tax inclusion.
- Residency: Chinese API agent-api.bigmodel.cn; precise execution region and residency controls unpublished.
- Mandatory model inference (bigmodel.cn pricing, CNY per M tokens, input / output / cache hit): GLM-5.3 8 / 28 / 2; GLM-5.3-FlashX 2 / 7 / 0.57; GLM-5.3-Flash list 0.8 / 2.8 / 0.23 with a limited-time 50% promo (0.4 / 1.4 / 0.115, no end date). Cache storage temporarily free. Not included in the runtime estimate.

## How this provider charges


As of 2026-09-28. Price basis: public list/reference unless labeled historical, illustrative, promo or unpriced.

API private beta with managed Agent/Environment/Session. Sandbox/runtime currently temporarily free, but model inference separately uses standard API balance (Coding Plan excluded). CPU/RAM dimensions and arbitrary 4/8 access not specified: zero marginal sandbox fee is NOT a free 4/8 VM quote. Filesystem checkpoint is opt-in; do not infer in-memory process resume from conversational session state.

## Currency, VAT and residency
- Native catalog: CNY. FX: 1 CNY = 0.149020326907 USD, ECB 2026-09-28 ([source](https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/index.cs.html)). USD-listed catalogs remain independent quotes.
- Tax: Unknown: reviewed pricing source does not explicitly establish VAT/tax inclusion.
- Residency/restrictions: Chinese API agent-api.bigmodel.cn; precise execution region and residency controls unpublished.

## Regimes
| Regime | When it applies | How billed | Native / USD numbers | Source |
|---|---|---|---|---|
| managed-agents-beta | Managed Agent sandbox, temporary free runtime fee | allocated wall time; None second increment | CNY None/vCPU-h + None/GiB-h → USD None + None; null means unknown | https://docs.bigmodel.cn/cn/managed-agents/overview.md |
| promo | Private beta runtime | See rule | CNY 0 sandbox/runtime fee temporarily. Do not price arbitrary requested VM shape as free. | https://docs.bigmodel.cn/cn/managed-agents/overview.md |
| inference | Every model call | See rule | Separate model tokens; no Coding Plan credit. Not represented by CPU-hour estimator. | https://docs.bigmodel.cn/cn/managed-agents/overview.md |
| checkpoint | Filesystem only | See rule | /mnt/session/outputs retained; /workspace and /tmp require x-checkpoint true. Docker CLI installed, daemon unavailable. | https://docs.bigmodel.cn/cn/managed-agents/overview.md |

## Gotchas
- Fixed CPU/RAM size, 50-concurrent access and 8-hour runtime eligibility unknown; no 4/8 comparison.
- Runtime promo end date unknown; model-token bill is mandatory but depends on selected model/tokens.
- Outbound network limited to HTTP/HTTPS ports 80/443; allowlists available.
- Unknowns remain null, not free; generic infrastructure modes are opt-in alt.
- Worked example assumes resources are released between active windows unless a retained-instance scenario is explicitly discussed.
- VAT basis: Unknown: reviewed pricing source does not explicitly establish VAT/tax inclusion.
- Residency: Chinese API agent-api.bigmodel.cn; precise execution region and residency controls unpublished.

## Required worked example
4 vCPU / 8 GiB requested; 50 simultaneous workers × 8 hours/day × 22 days = **8,800 instance-hours**, 1,100 daily sessions. CPU utilization 30% does not reduce allocated-resource charges. Snapshot retention is **50 GiB total for 730 hours**, not 50 GiB per worker; egress is **100 GiB total**. Root disk size was not specified, so root/volume costs are not silently assigned zero.
- managed-agents-beta: compute **unknown**; cannot produce a comparable quote.
- 50 GiB retained snapshots: **unknown/unverified feature or price**, not USD 0.
- 100 GiB egress: **unknown**, not USD 0.
- Full all-in bill: **null** where any required storage/network/fee or eligibility fact is unresolved. Quota raises may be necessary for 200 aggregate vCPU / 400 GiB RAM.

## Sources
- https://docs.bigmodel.cn/cn/managed-agents/overview.md
- https://docs.bigmodel.cn/cn/managed-agents/cloud-environment.md
- https://docs.bigmodel.cn/cn/managed-agents/sandbox-reference.md
- https://docs.bigmodel.cn/cn/managed-agents/faq.md
- https://docs.bigmodel.cn/cn/guide/start/pricing.md