TAONIC
Intelligence, routed.
Ask once. TAONIC identifies the required capability, routes the request through an available provider or Bittensor subnet, and returns the result with an inspectable record.
- VERIFIEDLive: routing, request records and measured provider cost
- PROPOSEDProposed: paid usage and a TAO reserve
- NOT DEPLOYEDTAON, the pool, the hook and the reserve
- ECONOMIC LAYER IN DEVELOPMENT
Reading which routes are connected…
- Question
- Capability
- Route
- Provider
- Answer
- Record
What an answer looks like
Example answerWhich Bittensor subnets were strongest for inference in September 2026?
The short answer
Subnet 64, Chutes, is the clearest live inference network on Bittensor: a public, OpenAI-compatible API with 14 open models running in trusted execution environments, served by miners registered on the subnet.
How the field looks
- Chutes (SN64): serverless inference for text, image and speech. Its primary text model alone served about 316,000 requests on 13 September.
- Verathos (SN96): newer, with proof metadata on every response. Access keys are created with a wallet signature.
- Targon (SN4): now rents GPUs and confidential machines instead of serving inference directly.
What "strongest" should mean
Requests actually served, models available, uptime, and how much of the subnet’s emissions reward real work. These move daily, so check them live before trusting any ranking.
Answered by Chutes · Bittensor subnet 64
- What this is
- An example of a TAONIC answer, written on 2026-09-14 from TAONIC’s own research into Bittensor inference routes. It is not a live response; ask your own question to get one.
- Provider
- Chutes, on Bittensor subnet 64. OpenRouter forwards the request to Chutes, whose miners on Bittensor subnet 64 serve the model. TAONIC is a customer of OpenRouter, which is a customer of Chutes; it does not run a miner or a validator.
- Integration
- Live text answers go through OpenRouter, pinned to Chutes on subnet 64, model google/gemma-4-31b-it. Every live answer carries its request id, timing and the provider’s own evidence.
- Sources
- Subnet 64 on taostats (opens in a new tab)Chutes platform index (opens in a new tab)Chutes public model list (opens in a new tab)Verathos API documentation (opens in a new tab)Targon documentation (opens in a new tab)
- Limits
- Rankings move daily. OpenRouter names the provider that served each answer, and TAONIC refuses the answer unless that provider is Chutes. Neither OpenRouter nor Chutes publishes which miner hotkey served a request, so no answer is attributed to a particular miner.
What you can ask
Seven kinds of question, each routed to the providers that do that work
Each row says whether it answers on this deployment now, and marks each step of its route ✓ when a connected route serves it. Available means it answers while part of its route waits; configuration required means the code is in place and its credential is not set, so nothing is sent.
Routes
Specialized providers and Bittensor subnets
TAONIC runs no subnet of its own. It routes each part of the work to a provider that does it, and a route counts as connected only after a real request has completed through it. States here come from the gateway and the durable records.
Inspect routes- Text generationChutes · SN64
- Text generationengy · SN53
- Text generationOpenRouter → Chutes · SN64
- Web searchDesearch · SN22
- Event forecastsNuminous · SN6
- Price forecastsSynth · SN50
- Image generationChutes · SN64
- Text to speechChutes · SN64
- Speech to textChutes · SN64
- EmbeddingsChutes · SN64
Future flywheelPROPOSED
Paid requests could one day build a TAO reserve
- Paid requests
- Provider cost deducted
- Eligible surplus
- TAO reserve growth
- Better capabilities
A proposal, in development. Charging is not enabled, so there is no paid usage, no surplus and no TAO accumulation today. Every value in this flow is a real zero. There is no token, pool, reserve or reward.