llm-prices

Website & Blog · GitHub · Changelog

A zero-dependency Python CLI and library for looking up and comparing LLM API costs across all major providers.

$ llm-prices top 10 --in 10000 --out 2000
Top 10 cheapest: 10,000 input / 2,000 output tokens

#    Model                  Provider            Input       Output        Total
-------------------------------------------------------------------------------
1    llama-3.1-8b-nb        Nebius          $0.000200    $0.000120    $0.000320
2    nova-micro-br          Bedrock         $0.000350    $0.000280    $0.000630
3    llama-3.1-8b           Groq            $0.000500    $0.000160    $0.000660
4    gemini-1.5-flash-8b    Google          $0.000375    $0.000300    $0.000675
5    command-r7b            Cohere          $0.000375    $0.000300    $0.000675
...

$ llm-prices compare gpt-4o claude-sonnet-4-6 gemini-2.5-flash gpt-5-nano --in 10000 --out 2000
Comparison: 10,000 input tokens, 2,000 output tokens

Model                Provider            Input       Output        Total
------------------------------------------------------------------------
gpt-5-nano           OpenAI          $0.000500    $0.000800    $0.001300
gemini-2.5-flash     Google          $0.003000    $0.005000    $0.008000 (6.2x)
gpt-4o               OpenAI            $0.0250      $0.0200    $0.0450 (34.6x)
claude-sonnet-4-6    Anthropic         $0.0300      $0.0300    $0.0600 (46.2x)

Cheapest: gpt-5-nano at $0.001300

Covers 196 models across 27 providers: OpenAI, Anthropic, Google, Mistral, Groq, Cohere, DeepSeek, xAI, Together AI, Fireworks AI, Perplexity, Cerebras, SambaNova, Amazon Bedrock, AI21 Labs, DeepInfra, Lambda AI, Novita AI, Nebius AI, Moonshot AI (Kimi K2), Hyperbolic, Crusoe, Qwen/Alibaba Cloud, Microsoft Azure, Inflection AI, NVIDIA NIM, NousResearch. No API key required — pricing data is baked in and updated with each release.

Also available as an MCP server — use llm-prices tools directly from Claude, Cursor, and other MCP-compatible AI assistants.

---

Install

pipx (recommended — installs globally, no venv required):

pipx install git+https://github.com/benbencodes/llm-prices

Homebrew (macOS/Linux):

brew tap benbencodes/tap
brew install llm-prices

pip (PyPI):

pip install llm-prices

From source:

git clone https://github.com/benbencodes/llm-prices
cd llm-prices
pip install -e .

Requires Python 3.8+. No other dependencies.

---

Usage

List all models

llm-prices list
llm-prices list --provider Anthropic
llm-prices list --search gemini --sort input
llm-prices list --json | jq '.[].model'

Export as Markdown table (for READMEs, docs, PRs):

llm-prices list --provider OpenAI --sort input --markdown

| Model        | Provider | Input/Mtok | Output/Mtok | Context  | Notes                     |
|--------------|----------|------------|-------------|----------|---------------------------|
| gpt-4.1-nano | OpenAI   | $0.1000    | $0.4000     | 1023k    | Fastest, cheapest GPT-4.1 |
| gpt-4o-mini  | OpenAI   | $0.1500    | $0.6000     | 128k     | Small, fast, cheap        |
| gpt-4.1-mini | OpenAI   | $0.4000    | $1.6000     | 1023k    | 1M context, cost-efficient|
| gpt-4o       | OpenAI   | $2.5000    | $10.0000    | 128k     | Latest multimodal flagship|
...

Export as CSV (for spreadsheets, databases):

llm-prices list --csv > llm_prices.csv

Calculate cost for a specific call

# 10,000 input tokens, 2,000 output tokens on GPT-4o
llm-prices calc gpt-4o --in 10000 --out 2000

# Model  : gpt-4o (OpenAI)
# Tokens : 10,000 in / 2,000 out
# Rate   : $2.5/Mtok in, $10.0/Mtok out
# Cost   : $0.0250 in + $0.0200 out = $0.0450 total

JSON output for scripting:

llm-prices calc claude-sonnet-4-6 --in 5000 --out 1000 --json

Compare models side-by-side

llm-prices compare gpt-4o claude-sonnet-4-6 gemini-2.5-pro qwen3-235b \
  --in 5000 --out 1000 --markdown

<!-- 5,000 input / 1,000 output tokens. Cheapest: qwen3-235b -->
| Model             | Provider  | Input     | Output    | Total            |
|-------------------|-----------|-----------|-----------|------------------|
| qwen3-235b        | Together  | $0.001000 | $0.000600 | $0.001600        |
| gemini-2.5-pro    | Google    | $0.006250 | $0.0100   | $0.0163 (10.2x)  |
| gpt-4o            | OpenAI    | $0.0125   | $0.0100   | $0.0225 (14.1x)  |
| claude-sonnet-4-6 | Anthropic | $0.0150   | $0.0150   | $0.0300 (18.8x)  |

Find the cheapest models for your workload

# Top 5 cheapest for 5k input / 1k output tokens
llm-prices top 5 --in 5000 --out 1000

Top 5 cheapest: 5,000 input / 1,000 output tokens

#    Model                Provider      Input       Output       Total
----------------------------------------------------------------------
1    llama-3.1-8b         Groq       $0.000250   $0.000080   $0.000330
2    gemini-1.5-flash-8b  Google     $0.000188   $0.000150   $0.000338
3    command-r7b          Cohere     $0.000188   $0.000150   $0.000338
4    qwen3.5-9b           Together   $0.000500   $0.000150   $0.000650
5    gemini-1.5-flash     Google     $0.000375   $0.000300   $0.000675

Filter to a single provider, or get a Markdown table:

llm-prices top 3 --provider Anthropic --in 2000 --out 800
llm-prices top 10 --in 5000 --out 1000 --markdown

How many calls fit in a budget?

# How many calls at 1k in / 500 out tokens fit in $1.00?
llm-prices budget 1.00 --in 1000 --out 500

# Filter to just Anthropic models
llm-prices budget 0.10 --provider Anthropic --in 5000 --out 2000

Use as a Python library

from llm_prices import calculate_cost, MODELS

result = calculate_cost("gpt-4o", input_tokens=10_000, output_tokens=2_000)
print(f"Total: ${result['total_cost_usd']:.4f}")

for name, info in MODELS.items():
    if info["provider"] == "Anthropic":
        print(name, info["input_per_mtok"], info["output_per_mtok"])

---

Providers & model count

Provider	Models	Notes
OpenAI	13	GPT-4o, GPT-4.1, o1, o3, o4
Anthropic	8	Claude 4, 3.7, 3.5, 3
Google	6	Gemini 2.5, 2.0, 1.5
Together AI	7	Qwen3, Kimi K2, Llama, DeepSeek
Fireworks	6	DeepSeek V4 Pro, V3, Kimi, Llama
Groq	9	Llama 4, Llama 3.x, Kimi K2, Qwen3 32B, gpt-oss 120B/20B
Mistral	12	Large 3, Magistral Medium/Small (reasoning), Medium 3, Small 3.2, Codestral, Devstral
Cohere	3	Command R+, R, R7B
Perplexity	4	Sonar, Sonar Pro, Reasoning, Deep Research
DeepSeek	3	chat (V3), V3.2, reasoner (R1)
xAI	8	Grok-4.3, Grok-4.20, Grok-4, Grok-4.1-fast, Grok-code-fast, Grok-3, Grok-3-mini, Grok-3-mini-fast
Cerebras	3	Llama 3.3 70B, Llama 3.1 8B, Qwen3 32B — ultra-fast silicon
SambaNova	5	Llama 4 Maverick, Llama 3.3 70B, DeepSeek-V3, MiniMax M2.5, Gemma 3 12B
Bedrock	5	Amazon Nova Micro/Lite/Pro/Premier/2-Lite — AWS-native foundation models
AI21	2	Jamba Mini 1.7, Jamba Large 1.7 — 256k ctx, hybrid SSM+Transformer
DeepInfra	4	Llama 4 Maverick (1M ctx!), Scout, DeepSeek-R1-0528, QwQ-32B
Lambda AI	4	Llama 4 Maverick ($0.05/Mtok — cheapest!), Scout, Llama 3.3 70B, DeepSeek-R1
Novita AI	4	Llama 4 Maverick (1M ctx), Scout, DeepSeek-R1-0528, Qwen3 235B
Nebius AI	5	Llama 3.1 8B ($0.02/Mtok — cheapest!), Llama 3.3 70B, Qwen3 235B (262k ctx), Nemotron 253B, DeepSeek-R1-0528

---

Pricing data

Prices are baked into the package at each release date and may drift behind provider changes. Check the sources for the latest. PRs updating llm_prices/data.py are welcome — please cite your source.

Sources

• OpenAI: https://openai.com/api/pricing/
• Anthropic: https://www.anthropic.com/pricing#anthropic-api
• Google: https://ai.google.dev/pricing
• Mistral: https://mistral.ai/technology/#pricing
• Groq: https://groq.com/pricing/
• Cohere: https://cohere.com/pricing
• DeepSeek: https://platform.deepseek.com/api-docs/pricing
• xAI: https://x.ai/api
• Together AI: https://docs.together.ai/docs/serverless-models
• Fireworks AI: https://docs.fireworks.ai/serverless/pricing
• Perplexity AI: https://docs.perplexity.ai/guides/pricing
• Cerebras: https://cerebras.ai/pricing
• SambaNova: https://api.sambanova.ai/v1/models (live API)
• Amazon Bedrock: https://aws.amazon.com/bedrock/pricing/
• AI21 Labs: https://www.ai21.com/pricing

---

Contributing

1. Fork the repo
2. Update llm_prices/data.py with new/corrected prices (cite your source)
3. Open a PR

---

MCP Server (for Claude, Cursor, and other AI assistants)

llm-prices ships with a built-in MCP server. Use it to query pricing data directly from any MCP-compatible AI assistant.

Install with MCP support

pip install "git+https://github.com/benbencodes/llm-prices[mcp]"

Configure Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):

{
  "mcpServers": {
    "llm-prices": {
      "command": "llm-prices-mcp"
    }
  }
}

Or for uvx users:

{
  "mcpServers": {
    "llm-prices": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/benbencodes/llm-prices[mcp]", "llm-prices-mcp"]
    }
  }
}

Available MCP tools

Tool	Description
get_model_pricing	Get pricing for a specific model
calculate_api_cost	Calculate exact cost for input+output tokens
compare_models	Compare cost of a workload across multiple models
find_cheapest_models	Find the N cheapest models for your workload
list_providers	List all 24 providers with min pricing
search_llm_models	Search models by name or filter by provider

---

Support this project

This tool is built and maintained by an AI agent. Donations go to the human operator's wallet. There is no promised return — this is a pure tip jar.

Prefer low-fee chains for small amounts (SOL, Base, Polygon, LTC, DOGE):

Chain	Address
SOL	kbghHYeBXr2AcYUyvkofHa9sArgkJcKBC6zZhSdao82
Base / ETH / EVM	0x310eEb225245D5A3e1773C5Def30Fe5d0289A1b3
LTC	ltc1q9fwegmfey7njksnmw8p787cz87l2lpf5372p2w
DOGE	DCHKeC2QQQSFVTA49gK44D1bfyv8QSnZyX
BTC	bc1qv0ny3c97lk80qv5v79f52w3hyaqq2ss0zdqp52
TRX / USDT-TRC20	TFaN8RPkgFkWjL5XHfJKRzyDQp2ECskQtH
XMR	4B3q6iZj8VJdZJLLWZggGSYsPWjMDhm8UJ6cfrkPbEHWCRqEvi1xyxtTbKZtbdeCLSdk17kvvgcyMVa2C59nkARfDgECSFd

---

License

MIT