NEURON
Docs ↗API

Tokenized inference · Base

1 NEURON = $1 of inference.

1 NEURON = $1 of inference on every model the gateway serves, at Venice's list price. The par door on Base mints one NEURON for one USDC, always.

NEURON … · 1 NEURON = $1 of inference

$1.00per NEURON at the par door
0%under Venice's list
0NEURON for sale under $1
Mint NEURON at $1.00Get an API key

Priced on the NEURON/KAI pool on Base through KAI/USDC; the par door at $1.00 is always open, so a NEURON never costs more than a dollar. The pools' own fees come on top, and the Buy card quotes the exact fill.

Models

One NEURON pays one list dollar on every model the gateway serves. Every price here is Venice's list. The gateway charges list, no markup.

NEURON back on a chat request an agent serves, when agents start selling their allowance

Text 127

Chat completions, streamed or not, priced per million tokens in and out.

ModelYour discountContextInput / 1MOutput / 1M
Gemini 3.6 Flashgemini-3-6-flash-1M$0.94$4.69
Gemini 3.7 Flashgemini-3-7-flash-1M$0.94$4.69
Gemini 3.8 Flashgemini-3-8-flash-1M$0.94$4.69
Gemini 3.5 Flash-Litegemini-3-5-flash-lite-1M$0.38$3.13
GLM 5.3z-ai-glm-5-3-1M$1.75$5.50
GLM 5.3 Flashz-ai-glm-5-3-flash-1.05M$0.15$0.50
GLM 5.2zai-org-glm-5-2-1M$1.40$4.40
GLM 5.1zai-org-glm-5-1-200K$1.54$4.84
GLM 5zai-org-glm-5-198K$1.00$3.20
GLM 5 Turboz-ai-glm-5-turbo-200K$1.20$4.00
GLM 5V Turboz-ai-glm-5v-turbo-200K$1.50$5.00
GLM 4.7 Flash Hereticolafangensan-glm-4.7-flash-heretic-200K$0.07$0.40

* Past 200K to 272K tokens of context, by model, these models bill a higher tier, up to 4 times the base input price. The gateway reserves at the dearest tier before a request runs, then charges exactly what Venice reports the request cost, at list, with no markup.

Images 40

A picture from a text prompt, priced per image; some models price by resolution and quality.

ModelYour discountPer image
Venice SD35venice-sd35-$0.01
Grok Imagine 2.0grok-imagine-image-2-0-from$0.051K $0.05 to $0.07 · 2K $0.07 to $0.10 · default $0.07
Grok Imagine High Quality (SOTA)grok-imagine-image-quality-from$0.061K $0.06 · 2K $0.09
Krea 2 Turbokrea-2-turbo-from$0.041K $0.04 · 2K $0.06
Flux 2 Proflux-2-pro-$0.03
Flux 2 Maxflux-2-max-$0.09

A tiered model is priced by the resolution and the quality /image/generate names, and two prices on one resolution are its low and its high quality. OpenAI's images/generations bills its default.

Speech 11

Text read aloud into an audio file, priced per 1,000 characters of input.

ModelYour discountVoicesPer 1,000 characters
Kokoro Text to Speechtts-kokoro-54$0.0035
Qwen 3 TTS 0.6Btts-qwen3-0-6b-9$0.0875
Qwen 3 TTS 1.7Btts-qwen3-1-7b-9$0.1125
xAI TTS v1tts-xai-v1-26$0.01875
Inworld TTS-1.5 Maxtts-inworld-1-5-max-14$0.0125
Chatterbox HD (Resemble AI)tts-chatterbox-hd-9$0.05

Embeddings 9

Text turned into a vector of numbers for search and retrieval, priced per million input tokens.

ModelYour discountVector lengthMax inputInput / 1M
BGE-M3text-embedding-bge-m3-1,0248K$0.15
BGE-EN-ICLtext-embedding-bge-en-icl-4,0968K$0.0125
Qwen3 Embedding 8Btext-embedding-qwen3-8b-4,09633K$0.0125
Qwen3 Embedding 0.6Btext-embedding-qwen3-0-6b-1,02433K$0.0125
Multilingual E5 Large Instructtext-embedding-multilingual-e5-large-instruct-1,0241K$0.0125
Text Embedding 3 Smalltext-embedding-3-small-1,5368K$0.025

Buy

Two doors, both on Base. The market sells under a dollar while its asks last; the par door is $1.00, always.

Buy on the market

Reading the pool…

The NEURON/KAI pool on Base; a USDC buy passes KAI's own USDC pool on the way, whose fees feed KAI and its backers. The fee row is read off both pools; your slippage is the only bound on the fill.

USDC

Mint at $1

One USDC in, one NEURON out, always. The dollar stays behind the NEURON until it is used for inference.

USDC
Balance 0 USDC

NeuronMinter on Base, at the same address on every chain, minting against that chain's own dollar. Pick the chain at the top of the page; a dollar paid in off Base waits in that chain's outbox until it is carried to Base.

Use

Activate, sign for a key, call any model above.

1

Activate

Activation burns the NEURON and credits its dollars to your address at the gateway, one balance whichever chain the NEURON was on: 1 NEURON becomes $1 of inference at list price. It is final; an activated dollar is spent on inference and on nothing else. Name an agent and its wallet gets the dollars instead, a gift nobody can take back.

NEURON
On Base -Activated -
2

Get a key

One signature from the wallet that activated, no gas. The key spends that address's activated dollars; a limit fences it inside them, for a bot or a teammate.

3

Use it

The gateway speaks OpenAI's chat completions, streamed or not, image generations, speech and embeddings. Point any OpenAI SDK at the base URL and pass the key.

Base URL
https://api.kairence.ai/api/v1
curl: chat
curl https://api.kairence.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $KAIRENCE_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "zai-org-glm-5", "max_tokens": 500, "messages": [{"role": "user", "content": "Hello"}]}'
curl: images
curl https://api.kairence.ai/api/v1/image/generate \
  -H "Authorization: Bearer $KAIRENCE_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "grok-imagine-image-2-0", "prompt": "A lighthouse at dawn", "resolution": "1K", "quality": "low"}'
curl: speech
curl https://api.kairence.ai/api/v1/audio/speech \
  -H "Authorization: Bearer $KAIRENCE_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "tts-kokoro", "input": "Hello from NEURON", "voice": "af_alloy"}' \
  --output speech.mp3
curl: embeddings
curl https://api.kairence.ai/api/v1/embeddings \
  -H "Authorization: Bearer $KAIRENCE_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "text-embedding-bge-m3", "input": "Hello"}'
JavaScript
import OpenAI from 'openai';
const ai = new OpenAI({baseURL: 'https://api.kairence.ai/api/v1', apiKey: process.env.KAIRENCE_KEY});
console.log((await ai.chat.completions.create({model: 'zai-org-glm-5', max_tokens: 500, messages: [{role: 'user', content: 'Hello'}]})).choices[0].message.content);
const imageBase64 = (await ai.images.generate({model: 'venice-sd35', prompt: 'A lighthouse at dawn', size: '1024x1024'})).data[0].b64_json;
const mp3 = Buffer.from(await (await ai.audio.speech.create({model: 'tts-kokoro', input: 'Hello from NEURON', voice: 'af_alloy'})).arrayBuffer());
const vector = (await ai.embeddings.create({model: 'text-embedding-bge-m3', input: 'Hello'})).data[0].embedding;

Name max_tokens: a request without it reserves the model's whole maximum output before it runs, and a small balance will not cover that.

Images: /image/generate bills the resolution and quality it names, one image a request. The SDK's images.generate is OpenAI's images/generations, which bills the model's default.

Backed

A dollar (USDC, or USDG on Robinhood Chain) stands behind every NEURON until it is activated: held by NeuronMinter on Base, waiting in a chain's outbox, on the road to Base, or kept by the road's relayers until the protocol repays it.

NEURON in circulation…
Dollars behind it…

Reading three chains…

In NeuronMinter on Base…
Waiting in the outboxes… USDG on Robinhood · … USDC on Arc…
On the road to Base0 of 8 legs not filled yet$0.00
The relayers' cut, repaid by hand each monthAcross keeps 25 bps of each leg; 8 legs since 2026-09-23$3.14
ChainNEURONDollarsContracts
BaseBase……NEURON 0x1010…0101 ↗NeuronMinter 0x06B4…1AE4 ↗
RobinhoodRobinhood……NEURON 0x1010…0101 ↗NeuronBackingOutbox 0x8B37…9157 ↗
ArcArc……NEURON 0x1010…0101 ↗NeuronBackingOutbox 0x8B37…9157 ↗

6 decimals on every side. Minting off Base puts the dollar in that chain's outbox, which sends it to NeuronMinter on Base in batches through Across; the relayer that fills a leg keeps up to 25 bps of it, and the protocol pays that cut back into the minter by hand each month. Activation always draws on Base.

Where it comes from

Sovereign agents earn NEURON from their pools' fees, 50% to the agent's backers and 50% to its Safe (all of it to the Safe while nobody backs the agent). They can also sell the inference their staked DIEM buys each day; when agents start selling their allowance, a request an agent serves pays that agent and pays you NEURON back. How it works ↗