Boundless CN (@boundless_cn) header
Boundless CN (@boundless_cn) profile picture

@BoundlessHQ 专为高速增长打造的推理引擎。极速部署开源模型,全程数据零留存

Joined July 2024
Photo by Boundless CN (@boundless_cn)Photo by Boundless CN (@boundless_cn)Photo by Boundless CN (@boundless_cn)Photo by Boundless CN (@boundless_cn)Photo by Boundless CN (@boundless_cn)Photo by Boundless CN (@boundless_cn)Photo by Boundless CN (@boundless_cn)Photo by Boundless CN (@boundless_cn)
Boundless CN retweeted
The power of inference is not measured by your ability to slop. It is measured by your ability to complete tasks - e2e, with high quality and amazing accuracy. Don't be a cog in a machine, be a machine! Be bold, be @BoundlessHQ
Getting from 0 to 80% of a task was always the easy part, long before AI, if you knew how to do the task. The last 20% is where the time goes: reviews, alignment, sign-off, guardrails, checklists. Agents can now carry that last stretch all the way to production. Every CTO, CIO and CPO should be building for 0 to 100, every single time.
2
7
491
72 小时肝出新官网,从敲定新定位到正式上线 → boundless.network 顺便重新认识一下现在的我们 → Boundless,专为高速增长打造的推理引擎。极速部署开源模型,全程数据零留存
Boundless: Inference for hypergrowth. New website → boundless.network
1
5
278
Boundless CN retweeted
New Boundless website is live: boundless.network Alongside it, updated positioning and messaging that reflects the last 2-3 weeks of customer convos. Boundless is inference for hypergrowth. We serve open models fast, with zero data retention, for companies growing faster than they can plan. I think of positioning as something that's always changing. It needs to be validated constantly, updated when the evidence changes, and owned by someone whose job is to do both. Never get married to it. A few weeks ago we placed a bet on "tailored inference." The idea: teams would want a partner who tunes models and serving to their traffic, not just another API. The market keeps you honest. Customers cared about a lot of things, but they REALLY cared about three: performance, zero data retention and reliability. Tailored inference is still one of our products. It just isn't the headline anymore. I knew we had to change, but in a way that still felt unique and ownable. Hence "inference for hypergrowth." We've come a long way in how we serve inference, and we realized we're the right partner for the fastest growing startups. We understand their needs, we can keep up with them, and we work the same way they do: a team of AI-pilled builders shipping faster than we can plan. This site went from new positioning to live in 72 hours. We understand our ICP because we ARE our ICP.
Boundless: Inference for hypergrowth. New website → boundless.network
7
8
1
29
2,439
周初许愿 Agent 数据库,周五交付 AmphoraDB 接锦鲤 🤩
This Friday, we run an experiment at Boundless. Announcing AmphoraDB: Serverless databases for agent swarms. No signup. No API key. Your wallet is the login. Paid per statement in stablecoins with MPP sessions on Tempo. Live on mainnet, now! 🧵
2
7
250
Boundless 现已支持精细化模型访问权限控制,帮团队建立清晰的算力边界 → 管理员可随时圈定团队基准模型池,把高阶 frontier 模型单独配发给研究组 → 开启限制后,所有新上线模型默认关闭,不经批准绝不产生一分钱意外账单 → 后台勾选,1 分钟全局同步所有现有 Key,越权请求直接零计费拦截 拒绝拿着大炮打蚊子,更防住新上线模型带来的账单刺客
Shared credits shouldn't mean surprise bills. Team admins on boundless can now choose exactly which models their team can use, give extra models to specific people, and keep new models off until they say so. Settings → members → model access
13
519
Boundless CN retweeted
Our 11/11 wish was 100% uptime and it is obviously a reality already. May your inference provider always be reliable, may it always remain @boundlessHQ.
1
3
11
1,619
Boundless CN retweeted
Boundless is still #1 provider by throughput for DeepSeek V4.1 Flash 1017 on Vercel AI Gateway → 313 tokens per second → ZDR, always xitter.cf/emilianobonassi/status…
I am hearing direct feedback from our customers about DeepSeek offensive cyber capabilities Clear this has to be reflected in benchmarks too At @BoundlessHQ we offer this and others at best throughput and reliability Join other teams and use it today 👇
7
2
17
2,999
调用顶尖模型能力而不必交出数据主权,开放模型是唯一解
Every prompt, attachment and Slack thread you send to an AI provider tells them something about your business CEO @sshankar chatted with @TheDryRunPod on why your inference provider should have no interest in learning your business, and how open models make it possible.
4
161
Boundless × @offensesh 正式打通 安全团队可以在一次性隔离沙箱中调用 Boundless 的 Abliterated 模型,让 Agent 自主执行授权安全测试,并带回调查结果与证据报告 红队真正需要的队友 沙盒实测:offense.sh 直连 API 申请:boundless.network/defensive-…
Boundless Defensive Access is now on @offensesh Security teams can run abliterated models inside Offense.sh's disposable sandboxes @boatdotdev for authorized security testing, starting with DeepSeek V4.1 Flash. Give your agent a security task & get back a report with findings and evidence.
2
2
9
703
Boundless CN retweeted
Toki.. 🫶 Named after the tokens every AI model streams.. fitting, since no one is using enough of them yet. ✊ Huge shoutout to our design folks + @agentsantic for bringing him to life. @BoundlessHQ is in good hands.
Intelligence deserves an identity. Say hi to Toki, our new mascot. 👋 You may have spotted him in our posts already, so it felt rude not to introduce him.
3
3
11
753
好乖;; 想rua 🥺🫳
Intelligence deserves an identity. Say hi to Toki, our new mascot. 👋 You may have spotted him in our posts already, so it felt rude not to introduce him.
6
74
Boundless CN retweeted
Boundless is now live on @useRouterPlus.
2
4
1
22
2,628
Boundless CN retweeted
A simple test from @sshankar on @TheDryRunPod: if AI keeps getting more useful and your bill doesn’t become longer, that means open models are doing the work. "It is hilarious how cheap open-source models are because it's free to run."
1
4
1
13
3,676
Boundless 最新推出区域推理路由 现在,你可以自主选择 AI 推理请求的运行区域,首批开放 US、EU 与 Asia 三大区域选择。结合 ZDR,提供更可控的推理服务 核心特性一览 → 按需路由:灵活指定 US/EU/Asia 推理节点,满足当地监管 →ZDR 原生架构:不持久化存储任何 Prompt 与响应数据,从底层杜绝隐私泄露 立即配置:inference.boundless.network/…
Introducing regional inference routing. Choose where your inference runs, starting with the US, the EU and Asia, with ZDR built in to meet data residency and compliance requirements.
1
4
335
重新定义 Vercel 上的推理极限,为极速交付而生
Boundless hit #1 for DeepSeek V4.1 Flash 1017 throughput on Vercel today. → 375 tokens per second → ZDR, always. Built for teams that ship fast.
4
95
Boundless 正式上线 Claude Fable 5.2 专为高强度工程打造,在代码生成、复杂推理与多模态视觉上迎来全面升级 ・1M 超长 Context Window + ZDR 架构,轻松吃下超大代码库与复杂工程文档 ・输入 $8/M Token,输出 $40/M Token,高阶算力性价比拉满 ・原生 MPP 支持:AI Agent 可直接通过稳定币自主结算 Fable 5.2 调用费用,真正打通 M2M的微支付闭环 欢迎直接体验:inference.boundless.network/…
Claude Fable 5.2 is now available on Boundless. Built for the hardest technical work, with improvements in coding, reasoning, and vision. 1M context window + ZDR.
5
120