Inference Exchange (@InferenceEx) header
Inference Exchange (@InferenceEx) profile picture

Run inference on a decentralized network in an open market of trusted users. Now entering private alpha, DM for access OCIP · Open Source · Building in Public

San Francisco, CA
Joined August 2026
inference metrics (e.g. TTFT, tok/s, etc). I understand this experiment is not simple to run, but a fully automated approach based on a scientific analysis of real data is perhaps the cleanest way to do it. From this, you can run your own experiments and collect data. N/N
1
1
31
different phones with common apps open, while measuring battery levels, temperature. Then you can run different inference requests repeatedly at a pace reflecting typical workflows, and see if there are any impacts to the device or the 4/N
1
1
26
across the different variables and measuring performance, are you able to get some deeper insights. I understand this is difficult to run in practice, but it can hopefully shed some light on the answer you may be looking for. For your use case, this could involve testing 3/N
1
1
25
team and can share some related practices. To get a fuller understanding, it requires a thorough experimental process. We would run several iPhones in temperature controlled environments, each with user profiles derived from telemetry data. Only after running the devices 2/N
1
1
23
Replying to @ayakthechad
Hi @ayakthechad, thanks for reaching out, great question. For us, we're not currently focused on edge devices, as our systems typically sit at a desk or in an office and send requests over the network. I (@jaredyu999) previously worked at @Apple with the battery data science 1/N
1
1
32
Replying to @thesimaximalist
Agreed! If you follow me back we can discuss further the private alpha. Seems like you’re currently closed to DMs.
12
Inference Exchange is entering private alpha. We're building an open and transparent marketplace for AI inference built on the Open Confidence Inference Protocol (OCIP). If you have Apple Silicon or use OpenAI's Python SDK feel free to reach out. Reply or DM for access.
1
1
1
102
Replying to @ChanleeMunger
Hi @ChanleeMunger, thanks for the question. We're looking to build a marketplace where both data center-level or local LLMs providers can dynamically offer to fulfill inference requests based on consumer demand. So once a user submits an API request it should be filled ASAP.
1
21
We've been building for a few months now, and we've been fortunate enough to secure some funding from family. Looking forward to pushing Inference Exchange to the next step.
1
1
158
Inference Exchange retweeted
reddit.com/r/LocalLLM/s/Ss2L… Hi all, sharing this Reddit thread on some research I’m doing into confidential computing for the purpose of local LLM inference with consumer hardware (e.g. Apple Silicon, Nvidia GPUs). Looking to learn security best practices from the community.
1
1
10
1,093