The models are open. Supply is what's scarce. 模型早已开放,稀缺的是供应。
Reliova is a token supply channel. We hold committed inference capacity for GLM, DeepSeek and Kimi, and deliver it on one OpenAI-compatible key — metered to the token, contracted for the long run. Reliova 是一条 Token 供应通道。我们持有 GLM、DeepSeek、Kimi 的承诺算力,通过一把兼容 OpenAI 的密钥交付——按 token 精确计量,以长期合约定价。
Buy capacity采购算力
Product teams running open models in production, on real monthly volume.在生产环境跑开源模型、有真实月度用量的产品团队。
Three ways to buy →三种买法 →Platforms & white-label平台与白牌
SaaS products, agencies and resellers who need inference behind their own brand.需要把推理能力装进自有品牌的 SaaS、外包公司与分销商。
Build on Reliova →基于 Reliova 构建 →Supply partners供应伙伴
Operators with clusters to fill, looking for steady, forecastable offtake.拥有集群产能、希望获得稳定可预测采购量的运营方。
Partner with us →成为供应伙伴 →Three ways to buy, one endpoint三种买法,同一个端点
Most workloads are a mix. Run your user-facing paths on demand, reserve throughput for the traffic you can forecast, and send everything that can wait to the batch tier.大多数业务是混合的:面向用户的链路按量走,能预测的流量用预留额度锁住,可以等的作业全部丢进批处理档。
On-demand按量计费
Pay per token as you go. Fund a balance, draw it down, top it up when you like. The right start for a team still finding its volume.按 token 实时扣费。充值、消耗、随时续充。用量还在爬坡的团队从这里开始最合适。
Reserved throughput预留吞吐
Lock a dedicated throughput block at a fixed monthly rate. Your cost stops moving with your traffic, and the headroom is held for you — with priority routing on top.以固定月费锁定一块专属吞吐额度。成本不再随流量波动,容量为您保留,并享有优先路由。
Batch tier批处理档
For work with no deadline: embeddings, backfills, evaluation runs, document pipelines. Submit the job, collect the output, pay materially less for it.给不赶时间的作业:向量嵌入、数据回填、评测跑批、文档流水线。提交任务、取回结果,价格显著更低。
Frontier open models, held capacity前沿开源模型,持有算力
The roster tracks the frontier as it moves. Need a model, a region or an SLA that isn't here? Tell us — we will source it or tell you straight that we can't.阵容随前沿模型持续更新。需要清单之外的模型、部署区域或 SLA?直接说,能配到我们就去配,配不到也会如实相告。
Sit behind your brand, not ours站在您的品牌背后
If you ship a product with AI inside it, inference is a supply problem you did not sign up to solve. Reliova runs underneath: your customers see your product, your pricing, your brand — and you see one contract and one invoice. 如果您的产品内置了 AI 能力,推理算力就是一个您本不想接手的供应链问题。Reliova 在底层运转:您的客户看到的是您的产品、您的定价、您的品牌,而您面对的只有一份合约、一张账单。
What you get underneath底层为您提供
Bring capacity. We bring the demand book.您提供算力,我们带来需求订单。
Idle accelerators earn nothing. Reliova aggregates enterprise demand into steady, forecastable offtake and settles it on one clean contract — so your clusters run loaded, including through the hours they would otherwise sit quiet. 闲置的加速卡不产生任何收益。Reliova 把企业需求聚合成稳定、可预测的持续采购量,以一份清晰的合约完成结算——让您的集群保持满载,包括那些原本安静的时段。
What a Reliova partnership looks like与 Reliova 合作意味着
Three steps to your first token三步跑通第一个 Token
Every account is shaped around the workload it actually runs, so the first conversation is a short technical one.每个账户都围绕真实业务负载来配置,所以第一次沟通是一场简短的技术对话。
Scope the workload明确用量场景
Models, monthly volume, latency and context needs. We size capacity against it and quote the mode that fits.模型、月度用量、延迟与上下文要求。我们据此配置算力,并给出最合适的那种买法与报价。
Test on a funded balance充值试跑验证
Fund a small test balance and run the real endpoint. Whatever you don't spend is refundable.先充一笔小额测试余额,直接跑真实端点,未消耗部分可退。
Scale on prepaid metering预付计量,规模扩展
Prepay, then draw down per token as you go — with a named contact and priority routing behind you.先预付,之后按 token 实时扣减,并配备专属对接人与优先路由。
Straight answers直接回答
What is Reliova?Reliova 是做什么的?
Reliova is a token supply channel. We hold committed inference capacity for frontier open-weight models and deliver it to companies through a single OpenAI-compatible API endpoint, billed per token.Reliova 是一条 Token 供应通道。我们持有前沿开源模型的承诺推理算力,通过一个兼容 OpenAI 的 API 端点交付给企业,按 token 计费。
Which models can I access?可以接入哪些模型?
GLM-5.3 and GLM-5.2 from Zhipu, DeepSeek V4 Pro and V4 Flash, and Kimi K3 from Moonshot AI. Every model on the roster supports a 1M-token context window.智谱的 GLM-5.3 与 GLM-5.2、DeepSeek V4 Pro 与 V4 Flash,以及月之暗面的 Kimi K3。阵容中每个模型都支持百万 token 上下文。
Do I have to change my code?需要改代码吗?
No. The endpoint is OpenAI-compatible, so you point your existing SDK at our base URL and use your Reliova key. Switching between models on the roster does not require code changes either.不需要。端点兼容 OpenAI 接口,把现有 SDK 的地址改成我们的,再换上 Reliova 的密钥即可。在阵容内切换模型同样不用改代码。
What throughput do I get?吞吐能力是多少?
Accounts run at 2,000 to 3,000 requests per minute and 10 million to 50 million tokens per minute, sized to the workload. Reserved throughput plans hold that headroom for you exclusively.账户的处理能力为每分钟 2,000 至 3,000 次请求、每分钟 1,000 万至 5,000 万 token,按业务负载配置。预留吞吐方案会为您独占保留这部分容量。
How does billing work?怎么计费?
Prepaid and metered. You fund a balance and it draws down per token as you use it. Test balances are refundable for whatever you do not spend, and there is a lower-priced batch tier for work with no latency deadline.预付加计量。充值后按 token 实时扣减,用多少扣多少。测试余额未消耗部分可退,另外还有一档价格更低的批处理档,适合没有延迟要求的作业。
Can I resell capacity to my own customers?可以转售给我自己的客户吗?
Yes. Resale is explicitly permitted. Platforms and agencies get per-tenant keys with separate quotas and usage records, and Reliova never appears to your end users.可以,合约明确允许转售。平台和外包公司会获得租户级密钥,额度与用量记录彼此独立,而 Reliova 不会出现在您的终端用户面前。
How long does it take to get started?多久能开始使用?
Send us your models and expected monthly volume, and we come back within two business days with a sandbox key and a quote. The sandbox key runs against the real endpoint with a capped balance.把所需模型与预计月度用量发给我们,两个工作日内回复测试密钥和报价。测试密钥直接对接真实端点,余额有上限。
Tell us what you're running告诉我们你在跑什么
Buying capacity, building on top of it, or supplying it — send the numbers and we'll come back within two business days with a key and a quote. 无论您是采购算力、基于算力做产品,还是提供算力,把数字发过来,两个工作日内我们回复密钥与报价。