Chutes is Bittensor subnet 64 (SN64), a project in the Inference and Compute lane. Serverless AI compute, real usage
One of Bittensor's most-used products serves trillions of AI tokens a month, yet its subnet emission share currently reads zero. That gap is the whole story.
It is an educational deep dive, not a recommendation to stake. For live price, pool depth, and performance, always open the live subnet page.
What is Chutes (SN64)?
Chutes is a serverless platform for running AI models. You pick a model, or bring your own code, send a request, and Chutes handles the servers, GPUs, scaling, and routing behind it. It is subnet 64 on Bittensor, built by Rayon Labs.
In plain English: It's like AWS Lambda, but for AI models, run across a decentralized network of independent GPU providers instead of one company's data centers.
Centralized reference points: Replicate, Modal, Together AI
How miners and validators work
Like every Bittensor subnet, SN64 is a two-sided market: miners produce work, validators score it, and emissions follow those weights under the network’s dTAO economics.
- Miners run "chutes" (containerized AI applications) on their own GPUs, decide which models to keep loaded and ready, and race to cold-start new ones when demand appears.
- Validators confirm that the GPUs miners claim are real and that the work was actually done, using a hardware-fingerprinting library called GraVal, then score miners on the compute they deliver.
Mechanism and design
A "chute" is an application, which the project's documentation compares to a single FastAPI app, that wraps a model and runs on top of a Docker image. Miners host these chutes on their GPUs and decide which models to keep "hot" in memory so they answer instantly. When a request arrives for a model nobody has loaded, miners compete to cold-start it first. Validators do not need expensive hardware to participate: they verify miners with GraVal, a library that checks VRAM capacity and runs device-seeded matrix multiplications to confirm a GPU is the one it claims to be. Chutes documents miner rewards as a weighted mix of compute delivered, invocation volume, the breadth of unique chutes served, and bounties for cold-starting new models.
The codebase backs up the activity story. The public SDK and CLI repository carries roughly 420 commits from 9 contributors, with the most recent commits landing in late May 2026, and Rayon Labs maintains separate miner and API repositories alongside it. The lead commits come from founder Jon Durbin. (Development activity here is from a live GitHub API check.)
Here is the tension worth understanding. Despite that usage, Chutes' subnet emission share reads 0% at the time of this snapshot. Under Bittensor's Taoflow model, live since November 2025, a subnet earns emissions based on its net TAO staking flows (staking minus unstaking), smoothed over a multi-week window. Subnets with positive net flows earn emissions; subnets where unstaking has outpaced staking earn zero. Chutes' trailing 7-day net flow was negative, on the order of 4,600 TAO out, which is why its emission share sits at zero right now. This is a statement about token flows, not about the product, and it can swing back as staking flows turn positive. It is a clean reminder that on Bittensor, how much a subnet ships and how much it currently emits are two different measurements, confirmed here across TaoSwap, TAO.app, and the IntoTAO mirror.…
Why the project exists
- The problem it solves: Running AI models in production means wrestling with GPU provisioning, scaling, and cost. Chutes lets a developer skip all of that and call a model over an API.
- The opportunity: Inference is where most real AI spending happens. A network that serves models cheaply and at scale is competing for actual usage, not just attention.
- The Bittensor advantage: Pooling GPUs from independent suppliers can undercut centralized providers on price, and the validation layer exists to keep those suppliers honest about the hardware they run.
- Traction signals: This is the unusual part. Chutes ranks first on the OpenRouter provider leaderboard and, by its own and third-party accounts, serves trillions of tokens a month across more than 800 GPUs. The usage is real and public.
What stakers should check before buying alpha
Subnet storytelling is cheap. On-chain and product checks are not. Use the same framework as our subnet metrics guide:
| Check | Why |
|---|---|
| Real product / docs / repo | Separates shipping teams from narrative-only slots |
| Pool depth (TAO) | Your exit door: thin pools amplify slippage |
| Net flows (7d / 30d) | Under flow-based emissions, capital votes with its feet |
| Emission share | Whether the market is funding this subnet right now |
| Ownership concentration | Large holders can move a small alpha market |
| Immunity / deregistration | Post-immunity weak slots can lose the race (see immunity guide) |
Pull the live numbers on Chutes · SN64 rather than trusting any static screenshot in a blog post.
Risk factors
- Emission share at zero: With net staking flows negative over the trailing window, the subnet currently earns no emissions under Taoflow. This can reverse if flows turn positive, but while it persists, miners and validators are not being subsidized by new emissions.
- Crowded market: Inference-as-a-service is competitive, both inside Bittensor and against well-capitalized centralized providers like Replicate, Modal, and Together AI. Price and reliability have to stay sharp.
- Operator-run validation surface: The about page describes a main validator that operates the platform on a set of H200 GPUs. A product that routes through an operator-run surface depends on that operator's uptime and neutrality.
- Corporate and token structure: Chutes is operated by a registered corporate entity (Chutes Global Corp, registered in Nevis), and exposure is tied to the subnet's alpha token, which carries the usual liquidity and price risk of a thinly traded market.
How this fits a TAO portfolio
Many holders keep a root vs subnet split: root for calmer TAO exposure, subnets like Chutes for thematic upside (and higher variance). If you use TaoScale, you can size SN64 from your wallet without handing over custody. See non-custodial staking.
- Live market: SN64 on TaoScale
- Browse and stake: /invest
- More subnet explainers: /blog/subnets
FAQ
What does Chutes do on Bittensor?
It's like AWS Lambda, but for AI models, run across a decentralized network of independent GPU providers instead of one company's data centers.
What is the SN64 alpha token?
When you stake TAO into subnet 64, you receive that subnet’s alpha. Its price versus TAO moves with the subnet’s pool (see what is dTAO).
How do I research Chutes before staking?
Read the mechanism above, open the official repo/docs when available, then verify pool depth, flows, and activity on SN64. Cross-check with the metrics checklist.
Is this financial advice?
No. TaoScale publishes educational explainers and live tools. Decisions to stake are yours alone.
Not financial advice. Educational overview of Chutes (SN64) compiled from public project materials and ecosystem documentation. Mechanisms and market structure change, so verify live data and official sources before acting.
