Knowledge base Q&A (RAG): what it costs and which AI to use
Staff or customers ask questions, the system fetches the right passages and the model answers from them. Companies putting internal docs, policies or product manuals behind a chat box.
The short answer
At 15,000 questions a month, we estimate $5.62 on DeepSeek V4 Flash 0423 (the cheapest model that cleared our quality bar), $51.65 on Kimi K2.6, and $404 on GPT-5.5. The flagship costs 7.8x the middle pick. Judged on summarising, extraction and reasoning tasks: the answer has to stay inside the passages it was given.
Simulate your knowledge base Q&A (RAG)
Pick a job, set how much of it you do, and see what every model would cost. Defaults are typical, not yours. Change them.
Prompt, history, documents
What the model writes
Advanced: caching, reasoning, retries, discounts
Share that can use cached-input prices where a model offers them
1 = none. 2 = hidden thinking doubles billed output
Extra calls on top of the base count
Batch jobs or committed spend. Check your provider; we assume none by default
Lowest benchmark score a model can have and still be picked. Higher is stricter.
Limit to certain providers
Untick any provider you cannot use, for data-residency, contract or policy reasons. Picks are recalculated.
15K questions a month, three ways to pay
Every model here cleared a score of 9.0 out of 10 on our summarise, extract json, reasoning benchmark tasks, ran without errors and fits a 32K-token window. 15 of 386 priced models qualify.
Cheapest model that clears the bar.
DeepSeek V4 Flash 0423
- Per question
- $0.0004
- Benchmark
- 9.67 / 10 · 9 prompts
- Context
- 1.05M tokens
Priced near the middle of the models that clear the bar; best score there.
Kimi K2.6
- Per question
- $0.0034
- Benchmark
- 9.67 / 10 · 9 prompts
- Context
- 262K tokens
What many people reach for first.
GPT-5.5
- Per question
- $0.027
- Benchmark
- 9.67 / 10 · 9 prompts
- Context
- 1.05M tokens
Choosing GPT-5.5 over Kimi K2.6 costs $352 more a month, $4,227 a year (7.8x the price).
Both scored 9.67 out of 10 on this kind of work. The extra money buys safety margin our benchmark cannot measure: it is 24 short prompts, and most passing models score near the top. If a wrong answer is costly for you, that margin can be worth it. Test on your own data before you decide.
If it grows
Pay-per-use cost scales in a straight line. Monthly cost at 1x, 3x, 10x and 100x today's volume.
| Model | 1x · 15K | 3x · 45K | 10x · 150K | 100x · 1.5M |
|---|---|---|---|---|
| DeepSeek V4 Flash 0423 | $5.62 | $16.87 | $56.24 | $562 |
| Kimi K2.6 | $51.65 | $155 | $517 | $5,165 |
| GPT-5.5 | $404 | $1,212 | $4,039 | $40,388 |
What a budget buys
- DeepSeek V4 Flash 0423267Kquestions a month, about 8,890 a day
- Kimi K2.629Kquestions a month, about 968 a day
- GPT-5.53,714questions a month, about 123 a day
Every model, priced for your workload
41 models, 15 hidden because their window is smaller than 32K tokens.
| Model | A month | Per question | Benchmark | Context |
|---|---|---|---|---|
| DeepSeek V4.1 Flash DeepSeekbelow the bar | $5.12 | $0.0003 | 8.89 (9) | 1.05M |
| DeepSeek V4 Flash 0423 DeepSeekBudget pick | $5.62 | $0.0004 | 9.67 (9) | 1.05M |
| Llama 4 Scout Metabelow the bar | $7.41 | $0.0005 | 8.11 (9) | 1.31M |
| Ministral 3 8B 2512 Mistral AIbelow the bar | $8.18 | $0.0005 | 6.56 (9) | 262K |
| Qwen3.8 Flash Alibababelow the bar | $9.88 | $0.0007 | 8.78 (9) | 1M |
| GLM 5.3 Flash Z Aibelow the bar | $10.32 | $0.0007 | 8.56 (9) | 1.05M |
| Mistral Small 4 Mistral AI | $10.54 | $0.0007 | 9.67 (9) | 262K |
| GLM 5.3 Z Ai | $12.05 | $0.0008 | 9.11 (9) | 1.05M |
| GPT-5.4 Nano OpenAI | $16.42 | $0.0011 | 9.67 (9) | 400K |
| Qwen3.8 27B Alibababelow the bar | $17.69 | $0.0012 | 7.22 (9) | 1.02M |
| MiniMax M2.5 Minimax | $18.35 | $0.0012 | 9.56 (9) | 205K |
| Muse Glimmer 30B Meta | $21.29 | $0.0014 | 9.00 (9) | 131K |
Benchmark = mean judge score out of 10 on the summarise, extract json, reasoning prompts of the Model Playground, with the number of prompts in brackets. Models marked n/a have no benchmark run, so we do not guess their quality. How the benchmark works.
What we assumed
- Fifteen thousand questions a month, around 700 a working day.
- Each question sends five retrieved passages plus instructions, about 4,500 tokens.
- Answers are short, about 350 tokens.
- 30% of input (instructions, shared passages) repeats between questions.
These are typical numbers, not measurements of your business. Edit them in the simulator above.
What drives the cost
Every question carries several retrieved passages as input, so cost follows how many passages you stuff into the prompt.
Watch out for
- Embedding and vector database costs are not included. This estimate covers the model call only.
- Retrieving fewer, better passages cuts cost more than switching models.
Questions people ask
How much will knowledge base Q&A (RAG) cost per month?
At 15,000 questions a month, we estimate $5.62 a month on DeepSeek V4 Flash 0423, the cheapest model that clears our quality bar for this job. Kimi K2.6 costs $51.65 and GPT-5.5 costs $404. These are estimates from the assumptions listed on this page, not a quote. Change the numbers to match your own volume.
What is the cheapest AI model for knowledge base Q&A (RAG)?
On current OpenRouter prices, DeepSeek V4 Flash 0423 from DeepSeek is the cheapest model that cleared our bar of 9.0 out of 10, scoring 9.7/10 on 9 benchmark prompts for this kind of work. The benchmark is small, so treat it as a shortlist and test on your own data before switching.
Is it worth paying for a flagship model for knowledge base Q&A (RAG)?
On our benchmark, GPT-5.5 scored 9.7/10 on this kind of work, and a model priced near the middle of the models that cleared the bar, Kimi K2.6, scored 9.7/10, while the flagship costs 7.8x as much. A flagship buys safety margin the benchmark cannot measure. If a wrong answer is expensive for you, that can be worth paying for. If not, you are probably overpaying.
What happens to the cost if knowledge base Q&A (RAG) grows ten times?
Pay-per-use pricing scales in a straight line, so ten times the questions is ten times the bill. Use the growth table on this page to see 3x, 10x and 100x for the same three models.
What makes knowledge base Q&A (RAG) expensive?
Every question carries several retrieved passages as input, so cost follows how many passages you stuff into the prompt.
Price a different job
- Customer support chatbotFrom $1.74 a month at 3,000 conversations
- Coding assistant for a teamFrom $4.12 a month at 4,400 requests
- Document summarisationFrom $1.86 a month at 2,000 documents
- Blog and content writingFrom $0.083 a month at 120 articles
- Invoice and form data extractionFrom $4.65 a month at 20K documents
- Ticket and content classificationFrom $11.30 a month at 500K items
- AI agent workflowsFrom $7.44 a month at 1,500 runs
- Sales email personalisationFrom $0.595 a month at 5,000 emails
- Meeting notes and summariesFrom $0.575 a month at 400 meetings
- Code review for pull requestsFrom $0.840 a month at 600 pull requests
- Personal AI assistant for a founderFrom $0.478 a month at 1,500 messages
- Text-to-SQL data analystFrom $2.54 a month at 6,000 queries