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 These figures reflect RECATOOLS research as of 12 Jul 2026 and may be out of date or incomplete.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
ASEAN Perspective
Volcengine Ark 火山方舟 (ByteDance) in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
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
