Home › Best › Cheapest LLM API
Cheapest LLM API in 2026
Ranked by blended list price — three input tokens to one output token, roughly the mix of a chat or retrieval workload. Cached input and batch discounts can pull the real bill lower; the calculator applies both. Six reseller-priced rows are excluded because their vendors no longer publish per-token rates.
Top 15 by blended price
| # | Model | Provider | Input $/M | Output $/M | Blended $/M | Context |
|---|---|---|---|---|---|---|
| 1 | GPT-6 Luna | OpenAI | $0.10 | $0.50 | $0.225 | 1,050,000 |
| 2 | Claude Haiku 5.5 | Anthropic | $0.10 | $0.50 | $0.225 | 200,000 |
| 3 | Mistral Small 4 | Mistral AI | $0.15 | $0.60 | $0.30 | 256,000 |
| 4 | GPT-5.6 Luna | OpenAI | $0.20 | $1.20 | $0.50 | 1,050,000 |
| 5 | GPT-5.4 nano | OpenAI | $0.20 | $1.25 | $0.512 | 400,000 |
| 6 | DeepSeek V4.1 Flash | DeepSeek | $0.30 | $1.20 | $0.60 | 1,100,000 |
| 7 | Gemini 3.1 Flash-Lite | $0.25 | $1.50 | $0.625 | 1,000,000 | |
| 8 | Gemini 3.5 Flash-Lite | $0.30 | $2.50 | $0.925 | 1,000,000 | |
| 9 | Mistral Large 4 | Mistral AI | $0.70 | $2.10 | $1.23 | 1,000,000 |
| 10 | Gemini 3.8 Flash | $0.75 | $3.75 | $1.69 | 1,000,000 | |
| 11 | Gemini 3.6 Flash | $0.75 | $3.75 | $1.69 | 1,000,000 | |
| 12 | GPT-5.4 mini | OpenAI | $0.75 | $4.50 | $1.88 | 400,000 |
| 13 | Grok 4.3 | xAI | $1.25 | $2.50 | $1.88 | 1,000,000 |
| 14 | GPT-6.1 Terra | OpenAI | $1 | $5 | $2.25 | 1,050,000 |
| 15 | Claude Haiku 4.5 | Anthropic | $1 | $5 | $2.25 | 200,000 |
Frequently asked
What is the cheapest LLM API in 2026?
By blended list price, the cheapest tier is GPT-6 Luna and Claude Haiku 5.5 at about $0.20 per million tokens, followed by Mistral Small 4 at $0.26 and DeepSeek V4.1 Flash at $0.48. Cached input and batch discounts can cut these by a further 50-90%.
Is the cheapest model always the best choice?
No. The cheapest models have smaller context windows and lower quality on hard tasks. Route by task: use a budget model for classification and extraction, and a frontier model only where the quality difference matters.
How is blended price calculated?
Blended price is (3 x input price + 1 x output price) / 4, which approximates a 3:1 input-to-output workload. Your own mix may differ, so check the calculator.
Browse all 45 models · Compare two models