
Inference at a global scale. Low latency. High throughput. Zero data retention. Built by the same team from @alchemy.
San Francisco
Joined October 2026
- Tweets125
- Following13
- Followers293
- Likes215
Pinned Tweet
⚗️ Alchemy Inference is live.
open-weight models. low latency. low costs.
Kimi, DeepSeek and GLM on an OpenAI-compatible API, with zero data retention 🔒
inference.alchemy.com
october was rough for AI infra.
@wiz_io counted 9+ platforms that disclosed security incidents, and the public reports all point to the same target: API keys stored in environment variables.
Digital assets have fought this playbook for years.
here's what to ask your provider 👇
questions worth asking any inference provider, including us:
🔑 who on your team can read my secrets. is it logged?
⏱️ can i use scoped, short-lived keys?
🗑️ is zero data retention the default?
📦 how do you vet your vendors and dependencies?
1
24
648
in digital assets, a leaked key means funds gone in minutes.
that's the bar @Alchemy has built infrastructure for for years, and it's the bar we're holding inference to.
trust center 👉 trust.alchemy.com
one week of @Alchemy Inference 🚀
we've spent it on calls with the teams who signed up.
the biggest buyers aren't enterprises running AI behind the scenes. they're teams whose product is the AI: chat apps and agents for support, coding, legal and bio.
when AI is the product, tokens are COGS. here's what they asked for:
💸 price: the #1 reason teams say they'd switch
⚡ latency: holds at peak, not just on average
🔒 data: zero retention, privacy you can verify
🧾 spend: per-agent budget caps, so it can't run away
we're onboarding teams by hand while we're in beta.
shipping AI to your own users on open-weight models?
request access 👉 inference.alchemy.com
1
70
Alchemy Inference retweeted
spent this year hearing the same thing: inference bills growing faster than users.
this week the waitlist proved it. onboarding as fast as we can without cutting corners. thank you for the patience 🙏
demand > expectations.
onboarding as fast as we can without compromising the experience we promised. if you’re on the waitlist, hang tight. we see you
trying to onboard everyone as fast as we can - apologies for the wait 😅🙏
honestly didn't expect this much demand and we want to make the experience great
1
13
440
we built Alchemy Inference around the four things that decide an inference provider:
💸 cost: lower costs, without compromising performance
⚡ latency: low, with high throughput
🛡️ reliability: from the team behind Alchemy's 99.99% uptime
🔒 data: zero data retention
all you have to do is:
🔌 point the SDK you're using at our endpoint
🔑 swap the key
🧠 pick a model
no new SDKs, no GPUs to manage ✅
it's that easy 👉 inference.alchemy.com
2/ we built Alchemy Inference around the four things that decide an inference provider:
💸 cost: lower costs, without compromising performance
⚡ latency: low, with high throughput
🛡️ reliability: from the team behind Alchemy's 99.99% uptime
🔒 data: zero data retention
1
11
3/ all you have to do is:
🔌 point the SDK you're using at our endpoint
🔑 swap the key
🧠 pick a model
no new SDKs, no GPUs to manage ✅
it's that easy 👉 inference.alchemy.com
8


