Models & pricing

uncensored models

22

context window

256K

from

$0.50 /1M

Hosted open-weight models

262,144-token context on every id; input and output billed separately; each id accepts what the model it imitates accepts

Model	Input	Structured output	In / cached / out per 1M	More
DeepSeek-V4-Pro-08133BIT-REAPDeepSeek · 1.6T MoE · 49B activeUncensored	Text in	JSON schema	$1.95 / $0.488 / $3.90	More
GLM-5.23BIT-REAPZ.ai · 744B MoE · 40B activeUncensored	Text in	JSON schema	$0.73 / $0.183 / $2.34	More
Kimi-K33BIT-REAPMoonshot AI · 2.8T MoE · 104B activeUncensored	Text + image in	JSON schema	$3.83 / $0.958 / $19.12	More
Nemotron3Ultra3BIT-REAPNVIDIA · 550B hybrid Mamba-Attention MoE · 55B activeUncensored	Text in	JSON schema	$0.75 / $0.188 / $3.30	More
MiniMax-M33BIT-REAPMiniMax · 428B MoE · 23B active · sparse attentionUncensored	Text + image in	JSON schema	$0.50 / $0.125 / $2.00	More
DeepSeek-V4-Flash-0731W4A16-AUTOROUND-REAPDeepSeek · 284B MoE · 13B activeUncensored	Text in	JSON schema	$0.50 / $0.125 / $2.00	More
Kimi-K2.6W4A16-AUTOROUND-REAPMoonshot AI · 1T MoE · 32B activeUncensored	Text + image in	JSON schema	$0.78 / $0.195 / $3.67	More
Laguna-S-2.1W4A16-AUTOROUND-REAPPoolside · 118B MoE · 8B activeUncensored	Text in	No structured output	$0.50 / $0.125 / $2.00	More
inklingW4A16-AUTOROUND-REAPThinking Machines · 975B MoE · 41B activeUncensored	Text + image in	JSON object	$1.42 / $0.355 / $6.07	More
MiMo-V2.5-ProW8A16Xiaomi · 1.02T MoE · 42B active · 8-bitUncensored	Text in	JSON schema	$0.50 / $0.125 / $2.00	More
MiMo-V2.5W8A16Xiaomi · 8-bit multimodal · W8A16Uncensored	Text + image in	JSON schema	$0.50 / $0.125 / $2.00	More
Hy3W8A16Tencent · 295B MoE · 21B active · 8-bitUncensored	Text in	JSON schema	$0.50 / $0.125 / $2.00	More

Shannon models

one flat rate for input and output

Model	Context	Input	Structured output	Price / 1M tokens
Shannon 1.6 Liteshannon-1.6-liteUncensored	192K	Text + image in	JSON schema	$3.90
Shannon 1.6 Proshannon-1.6-proUncensored	192K	Text + image in	JSON schema	$7.80
Shannon 2 Liteshannon-2-liteUncensored	192K	Text + image in	JSON schema	$3.90
Shannon 2 Proshannon-2-proUncensored	192K	Text + image in	JSON schema	$5.85
Shannon 3shannon-3Uncensored	192K	Text + image in	JSON schema	$3.35
Shannon 3 Proshannon-3-proUncensored	192K	Text + image in	JSON schema	$3.35
Shannon 3.1shannon-3.1Uncensored	192K	Text + image in	JSON schema	$3.35
Shannon 3.1 Proshannon-3.1-proUncensored	192K	Text + image in	JSON schema	$3.35
Shannon 3.5shannon-3.5Uncensored	192K	Text + image in	JSON schema	$3.35
Shannon Coder 1shannon-coder-1Uncensored	128K	Text in	JSON schema	$8.00

Jump to Quickstart Try Playground

How to read a row

Every model streams and can call tools. The other features of the API are described on the Capabilities page. Capabilities

Context window and output limits

• The context window covers everything the model reads for one answer: your messages, tool definitions and images.
• The Shannon 3 family fits a longer conversation by itself: it keeps the newest turns that fit into the window and answers.
• In a stream, a prompt that is too long is reported as an error frame with type invalid_request_error and code: "context_length_exceeded".
• max_tokens takes values from 1 to 65,536 on every model; max_output_tokens in GET /v1/models is the top of that range. The default is 4,096.
• On every model, max_tokens is the amount a request reserves from your balance.
• As the limit on the length of the answer, max_tokens is applied by the hosted open-weight models, shannon-1.6-lite, shannon-1.6-pro and shannon-coder-1.
• On the hosted open-weight models it limits the answer text. The reasoning before the answer is not counted against it, and a value below 256 acts as 256.
• A stream of shannon-1.6-lite, shannon-1.6-pro, shannon-coder-1 or the Shannon 3 family ends with finish_reason: "length" when the answer was cut at its limit.

Limits and balance

How a price is formed

• Prices are in USD per 1,000,000 tokens. Your balance is counted in tokens worth $5.00 per 1,000,000, so a model's price says how much balance its tokens take.
• Shannon models have one price. A call is charged usage.total_tokens at that price.
• Hosted open-weight models have three prices. Input that is read from the prompt cache is charged at the cached price, which is 25% of the input price rounded to $0.001. The rest of the input is charged at the input price, the output at the output price.
• Hosted open-weight models count tokens with each model's own tokenizer. GET /v1/models marks this as tokenizer: "original".
• shannon-coder-1 is counted in calls, not tokens, on /v1/chat/completions and /v1/messages. Its listed price applies on /v1/responses.

What a request reserves, worked examples of a charge and the Shannon Coder call allowance are on Limits and balance. Limits and balance

Model ids

Send one of these ids as model.

shannon-1.6-lite
shannon-1.6-pro
shannon-2-lite
shannon-2-pro
shannon-3
shannon-3-pro
shannon-3.1
shannon-3.1-pro
shannon-3.5
shannon-coder-1
DeepSeek-V4-Pro-0813-3BIT-REAP
GLM-5.2-3BIT-REAP
Kimi-K3-3BIT-REAP
Nemotron3Ultra-3BIT-REAP
MiniMax-M3-3BIT-REAP
DeepSeek-V4-Flash-0731-W4A16-AUTOROUND-REAP
Kimi-K2.6-W4A16-AUTOROUND-REAP
Laguna-S-2.1-W4A16-AUTOROUND-REAP
inkling-W4A16-AUTOROUND-REAP
MiMo-V2.5-Pro-W8A16
MiMo-V2.5-W8A16
Hy3-W8A16

• Upper and lower case do not matter, and spaces around the id are removed: SHANNON-3 and shannon-3 are the same model.
• The reply always carries the published id in model, whichever spelling you sent.
• A display name such as Shannon 3 is not an id.

Four shorter spellings are accepted as well.

An id that is not published is answered with status 400, type invalid_request_error and the message unknown model: <id> on all three endpoints. No other model answers in its place.

{
  "error": {
    "type": "invalid_request_error",
    "message": "unknown model: gpt-4o"
  }
}

{
  "type": "error",
  "error": {
    "type": "invalid_request_error",
    "message": "unknown model: gpt-4o"
  }
}

• Always send model. /v1/responses answers a request without it with 400. On /v1/chat/completions and /v1/messages such a request is answered by shannon-1.6-lite.
• POST /v1/tokenize and POST /v1/messages/count_tokens take the hosted open-weight ids.

List models

GET /v1/models returns the models of this page as JSON, with the same context windows, prices and capabilities. Use it to fill a model picker or to check a price in code.

GET https://api.shannon-ai.com/v1/models

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://api.shannon-ai.com/v1"
)

for model in client.models.list():
    print(model.id)

import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: 'YOUR_API_KEY',
  baseURL: 'https://api.shannon-ai.com/v1'
});

for await (const model of client.models.list()) {
  console.log(model.id);
}

curl "https://api.shannon-ai.com/v1/models"

The reply, shortened to one entry of each kind:

{
  "object": "list",
  "data": [
    {
      "id": "shannon-3",
      "object": "model",
      "created": 1700000000,
      "owned_by": "shannon-ai",
      "display_name": "Shannon 3",
      "context_window": 196608,
      "max_output_tokens": 65536,
      "pricing": {
        "input_per_million_usd": 3.35,
        "output_per_million_usd": 3.35,
        "currency": "USD"
      },
      "capabilities": {
        "tools": true,
        "vision": true,
        "reasoning": true,
        "reasoning_effort": false,
        "streaming": true,
        "json_schema": true
      },
      "endpoints": [
        "/v1/chat/completions",
        "/v1/messages",
        "/v1/responses"
      ]
    },
    {
      "id": "MiniMax-M3-3BIT-REAP",
      "object": "model",
      "created": 1700000000,
      "owned_by": "shannon-ai",
      "display_name": "MiniMax-M3 · 3BIT-REAP",
      "context_window": 262144,
      "max_output_tokens": 65536,
      "pricing": {
        "input_per_million_usd": 0.5,
        "cached_input_per_million_usd": 0.125,
        "output_per_million_usd": 2,
        "currency": "USD"
      },
      "tokenizer": "original",
      "capabilities": {
        "tools": true,
        "vision": true,
        "reasoning": true,
        "reasoning_effort": true,
        "streaming": true,
        "response_format": true,
        "json_schema": true,
        "prompt_caching": true
      },
      "endpoints": [
        "/v1/chat/completions",
        "/v1/messages",
        "/v1/responses",
        "/v1/messages/count_tokens",
        "/v1/tokenize"
      ]
    }
  ]
}

• GET /v1/models needs no key. The reply is the same for every caller and does not depend on your plan.
• The list holds the ids that can be called. A model that cannot be served at the moment is left out of it.
• There is no endpoint for one model: GET /v1/models/{id} answers 404 with type not_found_error.