Volcengine Ark 火山方舟 (ByteDance)
ByteDance's MaaS platform for Doubao, DeepSeek, GLM and Kimi
Overview
Volcengine Ark is ByteDance's model-as-a-service console for running Doubao alongside third-party models like DeepSeek and GLM, with fine-tuning, evaluation and a coding-agent subscription bundled in. Built for developers and enterprises already on Volcengine's cloud.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 3 Sep 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- Discounted Doubao/DeepSeek coding access
- Claude Code / Anthropic-compatible endpoint
- veCLI support
- Higher token allowance than Lite
- Priority access to newer Doubao models
- Same multi-model routing
- Per-model tiered token pricing
- Fine-tuning and evaluation tools
- Enterprise SLA via Volcengine sales
What you can produce with Volcengine Ark 火山方舟 (ByteDance)
- OpenAI- and Anthropic-compatible API endpoints
- Doubao model family (text, vision, speech, image, video)
- Third-party model hosting (DeepSeek, GLM, Kimi)
- Fine-tuning and evaluation tooling
- veCLI command-line tool
- Coding Plan subscription (Lite/Pro)
- Agent Plan credit-based subscription
Ark's biggest asset is breadth: first-party Doubao models sit next to DeepSeek, GLM and Kimi behind one OpenAI- and Anthropic-compatible API, so switching models is a config change rather than a rewrite. The Coding Plan is genuinely cheap - Lite runs 40 yuan a month, Pro 200 - and Claude Code users can point straight at the Anthropic-compatible endpoint. Pricing is granular to the point of being fiddly: rates shift by context-length tier, and the newer Agent Plan charges 2 to 2.5x the legacy token rate, more at long context, which is easy to miss until the bill arrives. Everything runs through a Volcengine cloud account, so onboarding is Chinese-first and there's no clear international data-residency story. Good fit for teams already inside the ByteDance/Volcengine stack who want one bill for many models; less appealing if you need a single flat per-seat price.
What people say
Ark is the model-serving layer of ByteDance's Volcengine cloud - inference, fine-tuning, evaluation and deployment for the Doubao family plus hosted third-party models including DeepSeek, GLM and Kimi, all behind OpenAI- and Anthropic-compatible endpoints. A dedicated command-line tool, veCLI, ships alongside it for scripted workflows, and Claude Code users can connect to Ark models natively through the Anthropic-compatible route.
Pricing is metered by the million tokens and varies a lot by model generation. Older Doubao-1.5-Pro runs about 0.4 yuan per thousand input tokens; newer Doubao-Seed-1.6-class models bill roughly 2.4 yuan per million tokens for input and 24 yuan for output at the 128-256K context tier, with output cost dropping to 2 yuan per million tokens when a request stays under 32K input and 200 output tokens. On top of raw API billing, Volcengine sells a Coding Plan (Lite at 40 yuan/month, discounted to 9.9 yuan for the first month; Pro at 200 yuan/month, discounted to 49.9) and, since May 2026, an Agent Plan pitched as the first bundled agent-credit subscription in the category - its per-token rate runs 2 to 2.5x the Coding Plan's, rising to 7.5x for context past 128K.
Chinese technical write-ups describe the console and MaaS tooling as mature and genuinely useful for production deployment, with the multi-model catalog cited as a real advantage over single-vendor platforms. The trade-off shows up in billing complexity - tiered token pricing that changes by context length and model family is easy to misjudge, and the promotional first-month rates on the Coding and Agent plans step up sharply after month one. Everything is provisioned through a Volcengine account, which keeps onboarding friction low for teams already in that ecosystem but adds a real barrier for anyone outside it - there's no dedicated international billing or data-residency track, and support and documentation default to Chinese first.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including Volcengine Ark 火山方舟 (ByteDance)'s official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Volcengine Ark 火山方舟 (ByteDance) unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Volcengine Ark 火山方舟 (ByteDance) directly →
Spotted something out of date? Suggest an update →
Alternatives to Volcengine Ark 火山方舟 (ByteDance)
More in LLMs & Chat
