What is the cheapest LLM for RAG (Retrieval Augmented Generation)?

The cheapest LLM suitable for RAG (Retrieval Augmented Generation) is Japanese Stable Diffusion XL via Fireworks AI at $0.0001 per million tokens (blended rate). There are 200 models in this category.

Should I use an open-source LLM for RAG (Retrieval Augmented Generation)?

There are 19 open-weight models available for RAG (Retrieval Augmented Generation). Open-weight LLMs like Llama and Mistral can be self-hosted for lower per-token costs at scale, but require infrastructure investment. For lower volumes, API-based providers often offer better value.

10 Best LLMs for RAG in 2026

The 10 best LLMs for RAG pipelines in 2026. Chat models ranked by context window, grounding accuracy, and price for Retrieval Augmented Generation across 23+ providers.

200

Models Compared

$0.0001

Cheapest

1140

Top ELO

ELO-Rated

What is the best LLM for RAG (Retrieval Augmented Generation)?

The best LLM for RAG (Retrieval Augmented Generation) is Qwen3 Embedding 0 6B Batch via Deepinfra at $0.0037 per million tokens. With an Arena ELO of 1140, it offers the best balance of quality and cost. 200 models are compared across all providers.

What Makes a Good LLM for RAG?

Large context window for retrieved passages

Faithful grounding in provided context

Low hallucination rate

Cost-effective for high query volumes

RAG Models — Ranked by Value

200 models

Sorted by value: models with higher Arena ELO and lower price rank first. Models without ELO scores are sorted by cheapest price.

#	Model	Provider	Input	Output	Quality	Released↓
49	GPT 5 Nano	Replicate	$0.050	$0.400	Strong	Aug 7
37	Qwen: Qwen3 Coder 30B A3B Instruct 30BOpen	Deepinfra	$0.070	$0.260	Good	Jul 14
18	DeepSeek R1 0528 Qwen3 8B	Novita AI	$0.060	$0.090	Frontier	May 28
1	Qwen3 Embedding 0 6B Batch	Deepinfra	$0.0050	Free	Good	Apr 28
3	Qwen3 Reranker 0 6B	Deepinfra	$0.010	Free	Good	Apr 28
4	Qwen3 Embedding 4B Batch	Deepinfra	$0.010	Free	Good	Apr 28
5	Qwen3 Embedding 0 6B	Deepinfra	$0.010	Free	Good	Apr 28
6	Qwen3 Embedding 4B	Deepinfra	$0.020	Free	Good	Apr 28
7	Qwen3 Reranker 4B	Deepinfra	$0.025	Free	Good	Apr 28
12	Qwen3 4B Fp8	Novita AI	$0.030	$0.030	Good	Apr 28
13	Qwen3 Embedding 8B Batch	Deepinfra	$0.040	Free	Good	Apr 28
15	Qwen3 Reranker 8B	Deepinfra	$0.050	Free	Good	Apr 28
16	Qwen3 Embedding 8B	Deepinfra	$0.050	Free	Good	Apr 28
22	Qwen3 8B Fp8	Novita AI	$0.035	$0.138	Good	Apr 28
25	Qwen3 235B A22b Instruct 2507	Deepinfra	$0.071	$0.100	Good	Apr 28
28	Qwen: Qwen3 235B A22B Instruct 2507 235BOpen	OpenRouter	$0.085	$0.120	Good	Apr 28
30	Qwen3 1.7b Fp8 Draft	Fireworks AI	$0.100	$0.100	Good	Apr 28
31	Qwen3 1.7b	Fireworks AI	$0.100	$0.100	Good	Apr 28
32	Qwen3 1.7b Fp8 Draft 40960	Fireworks AI	$0.100	$0.100	Good	Apr 28
33	Qwen3 1.7b Fp8 Draft 131072	Fireworks AI	$0.100	$0.100	Good	Apr 28
34	Qwen3 0.6b	Fireworks AI	$0.100	$0.100	Good	Apr 28
48	Qwen: Qwen3 14B 14BOpen	OpenRouter	$0.072	$0.288	Good	Apr 28
50	Qwen: Qwen3 30B A3B 30BOpen	Deepinfra	$0.080	$0.280	Good	Apr 28
38	Llama 4 Scout 17B 16e Instruct	Deepinfra	$0.080	$0.300	Frontier	Apr 5
44	Gemini 2.0 Flash Lite Preview 02 05	Google	$0.075	$0.300	Strong	Feb 5
45	Gemini 2.0 Flash Lite 001	Google	$0.075	$0.300	Strong	Feb 5
46	Gemini 2.0 Flash Lite	Google	$0.075	$0.300	Strong	Feb 5
26	DeepSeek R1 Distill Qwen 1.5b	Fireworks AI	$0.100	$0.100	Frontier	Jan 20
47	DeepSeek R1 Distill Qwen 14B	Novita AI	$0.150	$0.150	Frontier	Jan 20
2	Meta: Llama 3.2 1B Instruct 1BOpen	Deepinfra	$0.0050	$0.010	Budget	Sep 25
8	Meta: Llama 3.2 3B Instruct 3BOpen	Deepinfra	$0.020	$0.020	Good	Sep 25
35	accounts/fireworks/models/llama-v3p2-3b	Fireworks AI	$0.100	$0.100	Good	Sep 25
36	accounts/fireworks/models/llama-v3p2-1b	Fireworks AI	$0.100	$0.100	Budget	Sep 25
10	Meta Llama 3.1 8B Instruct Turbo	Deepinfra	$0.020	$0.030	Good	Jul 23
11	Meta Llama 3.1 8B Instruct 8B	Novita AI	$0.020	$0.050	Good	Jul 23
19	Llama 3.1 8B Instant	Groq	$0.050	$0.080	Good	Jul 23
27	accounts/fireworks/models/llama-v3p1-405b-instruct-long	Fireworks AI	$0.100	$0.100	Frontier	Jul 23
29	accounts/fireworks/models/llama-v3p1-70b-instruct-1b	Fireworks AI	$0.100	$0.100	Strong	Jul 23
17	Google: Gemma 2 9B 9B	Deepinfra	$0.030	$0.060	Good	Jun 27
39	Gemini 1.5 Flash	Deepinfra	$0.075	$0.300	Strong	May 14
40	Gemini 1.5 Flash 001	Google	$0.075	$0.300	Strong	May 14
41	Gemini 1.5 Flash 002	Google	$0.075	$0.300	Strong	May 14
42	Gemini 1.5 Flash Latest	Google	$0.075	$0.300	Strong	May 14
24	Gemma 7B IT	Groq	$0.050	$0.080	Budget	Feb 21
43	Mixtral 8x7b Instruct	Perplexity	$0.070	$0.280	Good	Dec 8
14	Mistral: Mistral 7B Instruct v0.3 7BOpen	Deepinfra	$0.028	$0.054	Good	Sep 27
20	Mistral: Mistral 7B Instruct v0.2 7BOpen	Deepinfra	$0.055	$0.055	Good	Sep 27
21	Mistral: Mistral 7B Instruct v0.1 7BOpen	Deepinfra	$0.055	$0.055	Good	Sep 27
9	meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo	Deepinfra	$0.020	$0.030	Good	Feb 23
23	Meta Llama 3.2 3B Instruct Turbo	Together AI	$0.060	$0.060	Good	Feb 23

Best LLM for Other Use Cases

Cheapest Models Coding Chatbots Summarization Creative Writing Data Extraction Customer Support Code Review Translation AI Agents

Frequently Asked Questions

The #1 LLM for RAG (Retrieval Augmented Generation) in 2026 is Qwen3 Embedding 0 6B Batch via Deepinfra at $0.0037 per million tokens. It has an Arena ELO of 1140, placing it among the highest-rated models. This top-10 ranking considers both quality (Arena ELO) and cost to find the best value across 23+ providers.