BullTensor
TAO $277.90
Market cap · 30d
$67.1M τ241K
2.09% / 24h ↘
24h volume
$0 —
Daily, 24 days
Supply
16.7%
3.5M ش
in circulation, out of 21M max
  • In circulation 3.5M ش · 16.7%
  • In the pool 2.92M ش · 13.9%
  • Not emitted yet 14.6M ش · 69.4%
Locks
$15.4M 23.0% of circulating supply
  • Owner lock 640K ش · 18.3%
  • Other holders (4) 164K ش · 4.7%
  • Not locked 2.7M ش · 77.0%
∞ Perpetual lock 803K ش · ↘ Decaying lock 1,043.22 ش

What is Chutes?

In short
  • Chutes is an alternative to ChatGPT's API: you send a question, an AI answers, and you pay per use.
  • The models are open AIs (DeepSeek, Qwen, Kimi…), running on the graphics cards of thousands of independent participants.
  • It's Bittensor's best-known inference subnet (SN64): many people use it without knowing it's a subnet.

When a chat app, a coding assistant or a game needs an AI, it sends questions and gets answers back. Chutes provides that service, like OpenAI does for ChatGPT, but with no single company behind it: the machines running the AI belong to participants of the Bittensor network.

It's Bittensor's most used decentralized AI inference project. It receives 8.97% of the network's rewards, about τ323.01 a day.

How it works

  1. 01
    You send a question

    From an app or in code, with a Chutes key. Tools built for OpenAI work as they are.

  2. 02
    A participant runs the AI

    The subnet's miners lend their graphics cards and run the requested model.

  3. 03
    The network checks and pays

    Validators measure the quality and reliability of answers; the best miners earn the most rewards.

Going further: the technical details

Chutes is a serverless inference platform for open-weight models (DeepSeek, Qwen, GLM, Kimi, Gemma), through an OpenAI-compatible API. Models are called "chutes".

Every tempo (360 blocks, about 72 minutes), SN64's alpha emissions are split: 41% to miners, 41% to validators and their stakers, 18% to the owner.

Create an API key on chutes.ai, then point any OpenAI-compatible client at the base URL:

curl https://llm.chutes.ai/v1/chat/completions \
  -H "Authorization: Bearer $CHUTES_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "deepseek-ai/DeepSeek-V4-Flash-0731-TEE", "messages": [{"role": "user", "content": "Hello"}]}'

Owner on-chain revenue

What the subnet's team earns: 18% of the tokens the subnet creates every day go to it. It's its on-chain salary, visible to everyone.

Per day (current)
$24,847.62
τ89.41 · 1,295.97 ش
Last 30 days
$765K
τ2,705.60
Last 90 days
$2.07M
τ8,876.18

Daily revenue in USD (owner emissions × alpha price × TAO/USD of the day).

Alpha locks

on-chain state 41 minutes ago

When a team locks its own tokens, it can't sell them: it's a sign of confidence in its project. Other holders can lock too, behind the team's key.

Locked by the owner
640K ش
18.3% of circulating supply · $12.3M
Type
∞ Perpetual
Owner emissions auto-locked
Total on the owner hotkey
804K ش
∞ 803K · ↘ 1.04K
Wallets locking
5
on the owner hotkey
ColdkeyHotkeyAlpha locked % circulatingTypeLast update
5FRYKh…UhAyS9 Owner owner hotkey 639,610.70 ش 18.27% ∞ Perpetual #9270043
5HMtFE…eS8CCx owner hotkey 163,249.41 ش 4.66% ∞ Perpetual #8865642
5Dnt7N…Mtqurq owner hotkey 692.38 ش 0.02% ↘ Decaying #9217746
5F2Hi9…qQgC1k owner hotkey 373.24 ش 0.01% ↘ Decaying #9217743
5GRJ1h…uWddmH owner hotkey 19.52 ش 0.00% ↘ Decaying #9218565
See the 20 latest lock events
  • Oct 12 22:12 Alpha lock 64.80 ش owner #9270043
  • Oct 12 21:00 Alpha lock 64.80 ش owner #9269683
  • Oct 12 19:48 Alpha lock 64.80 ش owner #9269323
  • Oct 12 18:36 Alpha lock 64.80 ش owner #9268963
  • Oct 12 17:24 Alpha lock 64.80 ش owner #9268603
  • Oct 12 16:12 Alpha lock 64.80 ش owner #9268243
  • Oct 12 15:00 Alpha lock 64.80 ش owner #9267883
  • Oct 12 13:48 Alpha lock 64.80 ش owner #9267523
  • Oct 12 12:36 Alpha lock 64.80 ش owner #9267163
  • Oct 12 11:24 Alpha lock 64.80 ش owner #9266803
  • Oct 12 10:12 Alpha lock 64.80 ش owner #9266443
  • Oct 12 09:00 Alpha lock 64.80 ش owner #9266083
  • Oct 12 07:48 Alpha lock 64.80 ش owner #9265723
  • Oct 12 06:36 Alpha lock 64.80 ش owner #9265363
  • Oct 12 05:24 Alpha lock 64.80 ش owner #9265003
  • Oct 12 04:12 Alpha lock 64.80 ش owner #9264643
  • Oct 12 03:00 Alpha lock 64.80 ش owner #9264283
  • Oct 12 01:48 Alpha lock 64.80 ش owner #9263923
  • Oct 12 00:36 Alpha lock 64.80 ش owner #9263563
  • Oct 11 23:24 Alpha lock 64.80 ش owner #9263203
How it's measured

Alpha locked via Lock / OwnerLock. A perpetual lock (∞) never unlocks; a decaying lock (↘) is released progressively (~130 days). An owner locking its own alpha signals confidence in the project.

Buybacks

Some teams use their revenue to buy back their own token and destroy it ("buyback & burn"): fewer tokens in circulation, like a company buying back its shares.

Alpha burned 30d
—
Buyback & burn
…
Manual burns
…
Emission burns
…
Burned by the owner (30d)
0.00 ش

The split between buyback & burn, manual burns and emission burns appears once the events of the period are indexed (history scan in progress).

60 days excluded: missing reading, or the counter moved more than the subnet can burn (e.g. runtime migration rewriting it).

TAO spent 30d
τ0
$0
TAO spent (all indexed)
τ0
Alpha burned 30d
0.00 ش
Operations
0

No buyback indexed yet.

How it's measured

On-chain events: AddStakeBurn (buyback & burn), alpha purchases (StakeAdded) and burns/recycles made by this subnet's declared treasury wallets.

Chutes models and prices (live)

● Chutes API responding · 134 ms · checked 17 minutes ago · 24h uptime 100.0%
ModelVendor Input $/1M tokensOutput $/1M tokensContext
Mistral-Nemo-Instruct-2407 TEE unsloth $0.0245 $0.0978 128K
Nemotron-3-Nano-Omni-30B TEE — $0.0245 $0.0978 128K
Qwen3.8-27B TEE Qwen $0.039 $1.25 256K
Qwen3-32B TEE Qwen $0.104 $0.416 40K
gemma-4-31B-turbo TEE google $0.12 $0.37 128K
Qwen3.6-27B TEE Qwen $0.25 $2 256K
Qwen3-235B-A22B-Thinking-2507 TEE Qwen $0.2989 $1.1957 256K
DeepSeek-V4-Flash-0731 TEE deepseek-ai $0.44 $1.32 1,024K
Kimi-K2.6 TEE moonshotai $0.45 $1.95 256K
Qwen3.5-397B-A17B TEE Qwen $0.45 $3 256K
GLM-5.1 TEE zai-org $0.98 $3.08 198K
DeepSeek-V3.2 TEE deepseek-ai $1 $1 128K
GLM-5.2 TEE zai-org $1.25 $3.95 1,024K
Kimi-K3 TEE moonshotai $1.85 $9.5 1,024K

Source: Chutes' public API llm.chutes.ai/v1/models, refreshed every 10 minutes. Off-chain data, unlike the rest of this page.

Chutes vs the centralized market

Chutes is cheaper than the median centralized provider on 9 of the 11 models also sold on the traditional market.

Model Chutes Market median vs median Official API Cheapest centralized
Qwen3-235B-A22B-Thinking-2507 $0.5231 $1.2125 (1) ▼ 57% cheaper — $1.2125 Venice, fp8
Qwen3.8-27B $0.3418 $0.7375 (19) ▼ 54% cheaper $0.9563 Alibaba $0.3675 Tenstorrent, fp4
Kimi-K2.6 $0.825 $1.4375 (16) ▼ 43% cheaper $1.7125 Moonshot AI $0.7792 Baidu, fp4
Kimi-K3 $3.7625 $5.8 (19) ▼ 35% cheaper $6 Moonshot AI $3.0475 InferenceNet, fp4
Qwen3.6-27B $0.6875 $1.04 (6) ▼ 34% cheaper $1.0125 Alibaba $0.8425 Reka
GLM-5.1 $1.505 $2.0913 (12) ▼ 28% cheaper $2.15 Z.AI $1.4814 Baidu, fp8
Qwen3.5-397B-A17B $1.0875 $1.3 (10) ▼ 16% cheaper $0.8775 Alibaba $0.8775 Alibaba
Qwen3-32B $0.182 $0.205 (3) ▼ 11% cheaper — $0.13 DeepInfra, fp8
GLM-5.2 $1.925 $1.9663 (25) ▼ 2% cheaper $2.15 Z.AI $0.258 Baidu, fp8
DeepSeek-V3.2 $1 $0.4375 (12) ▲ 129% pricier — $0.234 GMICloud, fp8
DeepSeek-V4-Flash-0731 $0.66 $0.231 (24) ▲ 186% pricier — $0.09 DeepInfra, fp8

Blended price in USD per million tokens (3 input tokens for 1 output token, a typical chat/agent mix). Market = centralized providers selling the same model on OpenRouter (number of providers in brackets), refreshed hourly; Bittensor-based providers are excluded. Very cheap offers often use stronger quantization (fp4): check quality before switching.

Chutes FAQ

What is Chutes AI?
Chutes is a serverless AI inference platform on Bittensor subnet 64 (SN64). It serves open-weight models (DeepSeek, Qwen, GLM, Kimi, Gemma…) through an OpenAI-style API, powered by GPUs run by Bittensor miners.
Is Chutes cheaper than centralized providers?
It depends on the model. On the 11 models also sold by centralized providers, Chutes is below the market median on 9: the biggest saving is on Qwen3-235B-A22B-Thinking-2507 (-57%), while DeepSeek-V4-Flash-0731 is 186% above the median. See the comparison table above.
Is Chutes AI free?
Not right now: all 14 models currently listed by Chutes are billed per token, starting at $0.02450 per million input tokens (Mistral-Nemo-Instruct-2407-TEE). Free models have been offered at times; the table above shows live prices.
Is Chutes down?
The Chutes API was responding when last checked, 17 minutes ago (134 ms), with 100.0% uptime over 24 hours.
Is the Chutes API OpenAI-compatible?
Yes, Chutes exposes an OpenAI-style API at https://llm.chutes.ai/v1, so OpenAI SDKs and tools that accept a custom base URL can use it with a Chutes API key.
What is the Chutes token?
Chutes' token is the alpha token of Bittensor subnet 64 (symbol ش). It trades against TAO in the subnet's on-chain pool: currently τ0.0689926 ($19.17), for a market cap of $67.1M.
What does TEE mean in Chutes model names?
Models suffixed “TEE” run inside a Trusted Execution Environment, confidential-computing hardware that isolates the model and your prompts from the machine's operator.