Cheapest LLM for RAG chatbots
In a retrieval-augmented (RAG) chatbot most of each prompt is retrieved text that changes on every question, so input price dominates and caching helps less than in an agent. Prices verified 2026-10-18.
Assumptions
- 1,500-token system prompt (cached), 4,000 tokens of retrieved chunks, 500 tokens of question and recent history
- 6,000 input and 400 output tokens per answer, 25% of the prompt served from cache
- 100,000 answers per month, standard (real-time) pricing
Change any of these in the calculator: every row below has a link that opens the same scenario for that model.
The short answer
- Cheapest flagship model: Mistral Large 4 at $4.00 (temporary sale price; regular $1.36 / $4.18 per 1M)
- Cheapest mid-tier model: Gemini 3.8 Flash at $4.99 (price rises 2027-01-01)
- Cheapest small model: Claude Haiku 5.5 at $0.67
Where the money goes
With 6,000 input tokens and 400 output tokens per answer, output accounts for 8% to 42% of the per-answer cost depending on the model's output/input price ratio. The retrieved chunks are the biggest lever you control.
Retrieving twice as much context (8,000 instead of 4,000 tokens per question):
- Mistral Large 4: $400 → $672 per month
- Gemini 3.8 Flash: $499 → $799 per month
- Claude Haiku 5.5: $66.50 → $107 per month
Better retrieval (fewer, more relevant chunks) usually saves more than switching models, and often improves answers too. The system prompt is the only stable prefix here, so caching saves less than in agent workloads; if your chatbot keeps a long conversation history, the history becomes cacheable and savings rise.
Full ranking (36 models)
| # | Model | Tier | Per 1,000 answers | Per month (100,000) | |
|---|---|---|---|---|---|
| 1 | Claude Haiku 5.5 Anthropic | small | $0.67 | $66.50 | Calculator |
| 2 | GPT-6 Luna OpenAI | small | $0.67 | $66.50 | Calculator |
| 3 | Ministral 3 8B Mistral AI | small | $0.76 | $75.75 | Calculator |
| 4 | Mistral Small 4 Mistral AI | small | $0.94 | $93.75 | Calculator |
| 5 | GPT-5.6 Luna OpenAI | small | $1.41 | $141 | Calculator |
| 6 | Codestral Mistral AI | small | $1.76 | $176 | Calculator |
| 7 | Gemini 3.1 Flash-Lite Google | small | $1.76 | $176 | Calculator |
| 8 | DeepSeek V4.1 Flash DeepSeek | small | $1.84 | $184 | Calculator |
| 9 | Gemini 3.5 Flash-Lite Google | small | $2.40 | $240 | Calculator |
| 10 | Gemini 2.5 Flash Google | small | $2.40 | $240 | Calculator |
| 11 | Mistral Large 4 Mistral AI sale price | flagship | $4.00 | $400 | Calculator |
| 12 | Gemini 3.8 Flash Google price changes 2027-01-01 | mid | $4.99 | $499 | Calculator |
| 13 | Gemini 3.6 Flash Google price changes 2027-01-01 | mid | $4.99 | $499 | Calculator |
| 14 | GPT-5.4 mini OpenAI | small | $5.29 | $529 | Calculator |
| 15 | Grok Build 0.1 xAI | small | $5.60 | $560 | Calculator |
| 16 | Claude Haiku 4.5 Anthropic | small | $6.65 | $665 | Calculator |
| 17 | Grok 4.20 (Reasoning) xAI | reasoning | $6.92 | $693 | Calculator |
| 18 | Grok 4.3 xAI | mid | $6.92 | $693 | Calculator |
| 19 | DeepSeek V4 Pro DeepSeek | flagship | $7.59 | $759 | Calculator |
| 20 | Mistral Medium 3.5 Mistral AI | mid | $9.97 | $998 | Calculator |
| 21 | Grok 4.7 xAI | flagship | $12.15 | $1,215 | Calculator |
| 22 | Gemini 4 Argon Google price changes 2026-12-01 | flagship | $13.15 | $1,315 | Calculator |
| 23 | Claude Sonnet 5.5 Anthropic | mid | $13.15 | $1,315 | Calculator |
| 24 | GPT-6.1 Sol OpenAI | mid | $13.15 | $1,315 | Calculator |
| 25 | Claude Sonnet 5 Anthropic | mid | $13.30 | $1,330 | Calculator |
| 26 | GPT-6 Sol OpenAI | mid | $13.30 | $1,330 | Calculator |
| 27 | Gemini 3.1 Pro (Preview) Google | flagship | $14.10 | $1,410 | Calculator |
| 28 | GPT-5.6 Terra OpenAI | mid | $14.10 | $1,410 | Calculator |
| 29 | Claude Sonnet 4.6 Anthropic | mid | $19.95 | $1,995 | Calculator |
| 30 | Claude Opus 5.5 Anthropic | flagship | $26.30 | $2,630 | Calculator |
| 31 | GPT-5.6 Sol OpenAI | flagship | $26.60 | $2,660 | Calculator |
| 32 | Claude Opus 5 Anthropic | flagship | $33.25 | $3,325 | Calculator |
| 33 | Claude Opus 4.8 Anthropic | flagship | $33.25 | $3,325 | Calculator |
| 34 | GPT-5.5 OpenAI | flagship | $35.25 | $3,525 | Calculator |
| 35 | Claude Fable 5.1 Anthropic | flagship | $65.38 | $6,538 | Calculator |
| 36 | GPT-6 Astra OpenAI | flagship | $66.50 | $6,650 | Calculator |
USD, standard list prices. Tags mark announced price changes and temporary sale prices.
Price is not quality
This page ranks by cost only. Models differ in accuracy, instruction following, tool use, latency, rate limits and data policies, and those differences often matter more than the price gap. Shortlist two or three models from the table, run them on a sample of your real inputs, and compare cost per successful result.
Related comparisons
More guides
- Cheapest LLM for coding agents
- Cheapest LLM for batch classification and extraction
- Long-context document analysis (400K tokens)
- Cheapest LLM for high-volume customer support
- Prompt caching explained, with savings
FAQ
What is the cheapest LLM for a RAG chatbot?
On this workload Claude Haiku 5.5 costs $0.67 per 1,000 answers. The cheapest mid-tier model is Gemini 3.8 Flash at $4.99. Test answer quality on your own documents before choosing on price.
Does a bigger context window lower RAG cost?
No. A larger window lets you send more retrieved text, but every token sent is billed. Cost tracks how much you retrieve, not how much the model could accept.