Use case ยท checked October 12, 2026
Cheapest LLM API for RAG search
Claude Haiku 5.5 and GPT-6 Luna tie as the cheapest at $37.50 a month. Our assumption: 100,000 searches a month, each sending about 2,500 tokens of retrieved passages and getting a 250-token answer.
| # | Model | Per month | Per request |
|---|---|---|---|
| 1 | Claude Haiku 5.5 | $37.50 | $0.00038 |
| 2 | GPT-6 Luna | $37.50 | $0.00038 |
| 3 | DeepSeek V4.1 Flash | $52.50 | $0.00052 |
| 4 | Gemini 3.1 Flash-Lite | $100 | $0.00100 |
| 5 | Gemini 3.5 Flash-Lite | $138 | $0.00137 |
| 6 | Mistral Large | $162 | $0.00162 |
| 7 | DeepSeek V4 Pro | $214 | $0.00215 |
| 8 | Gemini 3.8 Flash | $281 | $0.00281 |
| 9 | Gemini 3.6 Flash | $281 | $0.00281 |
| 10 | Grok Build 0.1 | $300 | $0.00300 |
| 11 | Claude Haiku 4.5 | $375 | $0.00375 |
| 12 | Grok 4.3 | $375 | $0.00375 |
| 13 | Grok 4.7 | $650 | $0.00650 |
| 14 | Claude Sonnet 5.5 | $750 | $0.00750 |
| 15 | GPT-6.1 Sol | $750 | $0.00750 |
| 16 | GPT-5.3 Codex | $788 | $0.00788 |
| 17 | Gemini 3.1 Pro | $800 | $0.00800 |
| 18 | Claude Opus 5.5 | $1,500 | $0.01500 |
| 19 | Claude Fable 5.1 | $3,750 | $0.03750 |
| 20 | GPT-6 Astra | $3,750 | $0.03750 |
Change the assumptions
Start from a use case
1,000 tokens is about 750 English words. Input includes your system prompt, chat history and any retrieved documents.
Claude Haiku 5.5
AnthropicCheapest
$37.50 /mo
GPT-6 Luna
OpenAICheapest
$37.50 /mo
DeepSeek V4.1 Flash
DeepSeek
$52.50 /mo
Gemini 3.1 Flash-Lite
Google
$100 /mo
Gemini 3.5 Flash-Lite
Google
$138 /mo
Mistral Large
Mistral
$162 /mo
DeepSeek V4 Pro
DeepSeek
$214 /mo
Gemini 3.8 Flash
Google
$281 /mo
Gemini 3.6 Flash
Google
$281 /mo
$300 /mo
Claude Haiku 4.5
Anthropic
$375 /mo
Grok 4.3
xAI
$375 /mo
Grok 4.7
xAI
$650 /mo
Claude Sonnet 5.5
Anthropic
$750 /mo
GPT-6.1 Sol
OpenAI
$750 /mo
GPT-5.3 Codex
OpenAI
$788 /mo
Gemini 3.1 Pro
Google
$800 /mo
Claude Opus 5.5
Anthropic
$1,500 /mo
Claude Fable 5.1
Anthropic
$3,750 /mo
GPT-6 Astra
OpenAI
$3,750 /mo