Conifer (YC S26)

Stop paying cloud prices for every AI request. We route 80% of your queries to your own hardware at no cost, using the cloud only for your most complex tasks. Available now @ conifer.build.

Excited to share that @coniferbuild is part of the @ycombinator S26 batch. Looking forward to getting to work with @gustaf! Token costs are eye-watering. @charles_v11 and I burned through $13,000 in Claude credits in just five days, and that's a fraction of what every company transitioning to Al is facing. But it's not just the cost. Al usage today is fragmented: five different models, three subscriptions, three API dashboards, and a drawer full of API keys. Every team is flipping between tabs, juggling providers, and paying full price for all of them. Conifer replaces all of it with one interface. We're the inference gateway that routes every Al query to the cheapest model that can actually do the job. Built on top of our Typhoon engine, which allows local models to run faster and punch far above their weight, Conifer only reaches for the cloud on your hardest tasks. The result is a token bill >80% less. We launch tomorrow, July 7th, on conifer.build If you're part of a company spending thousands on Al every day, been holding off on the switch because of cost, or someone just trying to decrease their spend, we'd love to hear from you. Email contact@conifer.build or reach out here on X.

Llama.cpp is excellent. We made it faster anyway. typhoon vs llama.cpp: • +51% on LFM2-350M, faster on all 14 models • 1.03× → 1.73× as context scales 512 → 32k • the longer the prompt, the bigger the win Full breakdown: conifer.build/docs

Conifer is launched! Local AI runtime + IDE. Own engine (beats llama.cpp on a few benchmarks), real coding IDE on top, OS-sandboxed local agent. No cloud, no Docker, nothing leaves your machine. Free and open source, available now: conifer.build

That's Ministral 3 14B, reasoning variant released in December 2025. Larger specialized models score higher; that isn't the point. The point is that a year ago this sentence wasn't writable. For most well-scoped tasks, the gap between an open model and a frontier one has narrowed to where capability is no longer what decides it. So for most of that work, you're paying a frontier API for capability you don't need. What decides it now is friction - getting these models to run well, locally, and reliably. That's the part Conifer is building: model management, hardware-aware execution, and the tools you already use point at your own machine instead of a hosted endpoint. Try it on June 1st conifer.build

You're wasting API tokens. A 14B model that runs on a laptop now solves 85% of AIME 2025

The standard is walking toward us. The models are walking toward us. The NSA just published the threat model. We don't have to argue our architecture anymore. We were early. Now everyone else is catching up.

Supply side is getting better underneath us at the same time. Qwen 3.5 shipped a full on-device tier this week. 0.8B, 2B, 4B, 9B. Apache-2.0. Built for edge. Fresh, permissive, actually good small models. Our runtime gets to be the best home for them.

The NSA just described our product without knowing it. Fort Meade dropped "Model Context Protocol: Security Design Considerations" this weekend. Every risk they name is one Typhoon was built to eliminate.

Signup for the waitlist here: conifer.build/feedback

No internet connection? No problem. Work anywhere securely with Conifer. Sign up for our waitlist today, we’re only taking 100 people from the waitlist! (Signup link in the replies)

The things you actually want an AI assistant to do? Read your email. Watch your files. Notice when something changes. That's also the list of things you'd never hand to a cloud model. We're building @coniferbuild. Beta drops June 1st, free and open source, join the waitlist now.

Hi guys, during the school year I've been focusing a lot of finance related projects, like marketsandmergers, and as of recent, alphabrief. However, this past couple of weeks, I've been involved in a new project called Conifer. The whole idea behind conifer is trying to optimize local ai. As of right now, most of the pain of running models locally isn't the model itself, it's everything around it: setup, storage, quantization, memory, and scheduling work across whatever hardware you've got. Conifer handles that layer so local inference runs fast and stays private, instead of being a fragile fallback to the cloud. We've got funding to build it, and right now we're already beating llama.cpp on some benchmarks, with more improvements coming as we prepare for launch. We're launching our beta on June 1st for our waitlist users. It'll be completely free and open source, so you'll be able to install it, point it at a model, and see what it does for yourself. If you're interested, we'd love for you to join the waitlist. Links will be in the comments if you guys are interested! This will be free + open source, and as of right now, we are only taking 100 people off of the waitlist.

Every time OpenAI posts "investigating elevated latency," it's a reminder that the runtime you depend on lives somewhere else. We're building conifer so it doesn't have to. Local, fast, no status page to refresh. 9to5mac.com/2026/05/27/ope…

You can sign up for the waitlist here: conifer.build/feedback

Conifer is an open-source runtime that makes local AI actually fast, private, and reliable. Here's the problem we kept running into: running AI models on your own machine should be the obvious choice for anything private, your code, your documents, your data. But anyone who's tried knows the truth: the hard part isn't the model, it's everything around it. Setup. Storage. Quantization. Memory management. Getting it to actually use your GPU properly. Splitting a model across multiple cards without it falling over. Keeping it stable when it runs for hours instead of minutes. Most people hit that wall and go back to the cloud. Which means their data leaves their machine, their costs scale forever, and their tool breaks the day someone else's servers have a bad day. Conifer is the layer that handles all of it. It manages the model, setup, storage, quantization, and memory, and schedules the work across whatever hardware you actually have, so local inference just runs. The goal is simple: local AI that's genuinely competitive with the cloud, not a slower fallback you settle for. And we're getting there. In early benchmarks Conifer already beats llama.cpp on decode for some models (1.24x on Qwen, 1.13x on TinyLlama) and MLX on prefill (1.26x on Qwen). It's not a clean sweep yet, there are places we're still behind and tuning hard, but it's genuinely competitive, and the numbers are climbing before launch. Conifer launches June 1st. Completely free and open source. Install it, point it at a model, and see what it does for yourself, no signup, no trial, no catch. We're opening beta access to the first 100 people on the waitlist before then. If you run models locally, or you've wanted to but the setup beats you, we'd love for you to try it and tell us what breaks. Link for signups will be in the replies.