LLM API pricing / Long prompts
The cheapest LLM APIs when the prompt is long
One request with a 300,000-token prompt and a 2,000-token answer costs $0.0308 on Gemini 2.5 Flash-Lite and $6.15 on GPT-6 Astra, across the 50 current models with a verified price that can take it. 18 of them charge a higher rate above a prompt length and are priced at that rate here. Prices verified between September 11, 2026 and October 1, 2026.
How the cost is computed
cost = 300,000 / 1,000,000 × input price + 2,000 / 1,000,000 × output price, with each model's price per 1M tokens. When a model records a long-prompt tier whose condition a 300,000-token prompt meets, the tier's input and output prices replace the base prices for the whole request; with several tiers, the highest one the prompt reaches. Cached input, batch and other price lists are not used: this is one uncached request at the list price. A model whose context window is smaller than 302,000 tokens cannot take the request and is listed apart; so is a model whose context window this site has not recorded.
Ranking
| # | Model | Cost of the request | Input / output $ per 1M applied | Context |
|---|---|---|---|---|
| 1 | Gemini 2.5 Flash-Lite Google | $0.0308 | $0.10 / $0.40 | 1,048,576 |
| 2 | Qwen3.8-Flash Alibaba | $0.0459 | $0.15 / $0.47 | 1,000,000 |
| 2 | Qwen3.8-Omni-Flash Alibaba | $0.0459 | $0.15 / $0.47 | 1,000,000 |
| 4 | GPT-6 Luna OpenAI | $0.0615 | $0.20 / $0.75 · tier over 272,000 tokens | 1,050,000 |
| 5 | Qwen3.7-Flash Alibaba | $0.0616 | $0.20 / $0.80 · tier over 256,000 tokens | 1,000,000 |
| 6 | DeepSeek V4.1 Flash DeepSeek | $0.0924 | $0.30 / $1.20 | 1,000,000 |
| 7 | Gemini 2.5 Flash Google | $0.0950 | $0.30 / $2.50 | 1,048,576 |
| 7 | Gemini 3.5 Flash-Lite Google | $0.0950 | $0.30 / $2.50 | 1,048,576 |
| 9 | GPT-5.6 Luna OpenAI | $0.124 | $0.40 / $1.80 · tier over 272,000 tokens | 1,050,000 |
| 10 | Gemini 3.6 Flash Google promo | $0.233 | $0.75 / $3.75 | 1,048,576 |
| 10 | Gemini 3.7 Flash Google promo | $0.233 | $0.75 / $3.75 | 1,048,576 |
| 10 | Gemini 3.8 Flash Google promo | $0.233 | $0.75 / $3.75 | 1,048,576 |
| 13 | Qwen3.7-Plus Alibaba | $0.370 | $1.20 / $4.80 · tier over 256,000 tokens | 1,000,000 |
| 14 | Muse Spark 1.2 Meta | $0.384 | $1.25 / $4.25 | 1,000,000 |
| 14 | Muse Spark 1.3 Meta | $0.384 | $1.25 / $4.25 | 1,000,000 |
| 16 | DeepSeek V4 Pro DeepSeek | $0.404 | $1.32 / $3.96 | 1,000,000 |
| 17 | GLM 5.2 Mistral | $0.429 | $1.40 / $4.40 | 1,000,000 |
| 17 | GLM 5.3 Mistral | $0.429 | $1.40 / $4.40 | 1,000,000 |
| 19 | Gemini 3.5 Flash Google | $0.468 | $1.50 / $9.00 | 1,048,576 |
| 20 | GPT-5.3 Codex OpenAI | $0.553 | $1.75 / $14.00 | 400,000 |
| 21 | Qwen3.8-Max Alibaba | $0.612 | $2.00 / $6.00 | 1,000,000 |
| 22 | Claude Sonnet 5 Anthropic | $0.620 | $2.00 / $10.00 | 1,000,000 |
| 22 | Claude Sonnet 5.5 Anthropic | $0.620 | $2.00 / $10.00 | 1,000,000 |
| 24 | Grok 4.20 (non-reasoning) xAI | $0.760 | $2.50 / $5.00 · tier of 200,000 tokens or more | 1,000,000 |
| 24 | Grok 4.20 (reasoning) xAI | $0.760 | $2.50 / $5.00 · tier of 200,000 tokens or more | 1,000,000 |
| 24 | Grok 4.20 Multi-agent xAI | $0.760 | $2.50 / $5.00 · tier of 200,000 tokens or more | 1,000,000 |
| 24 | Grok 4.3 xAI | $0.760 | $2.50 / $5.00 · tier of 200,000 tokens or more | 1,000,000 |
| 28 | Qwen3.7-Max Alibaba | $0.765 | $2.50 / $7.50 | 1,000,000 |
| 29 | Gemini 2.5 Pro Google | $0.780 | $2.50 / $15.00 · tier over 200,000 tokens | 1,048,576 |
| 30 | Claude Sonnet 4.6 Anthropic | $0.930 | $3.00 / $15.00 | 1,000,000 |
| 31 | Grok 4.5 xAI | $1.22 | $4.00 / $12.00 · tier of 200,000 tokens or more | 500,000 |
| 31 | Grok 4.6 xAI | $1.22 | $4.00 / $12.00 · tier of 200,000 tokens or more | 500,000 |
| 31 | Grok 4.7 xAI | $1.22 | $4.00 / $12.00 · tier of 200,000 tokens or more | 500,000 |
| 34 | GPT-6 Sol OpenAI | $1.23 | $4.00 / $15.00 · tier over 272,000 tokens | 1,050,000 |
| 34 | GPT-6.1 Sol OpenAI | $1.23 | $4.00 / $15.00 · tier over 272,000 tokens | 1,050,000 |
| 36 | Gemini 3.1 Pro Preview Google | $1.24 | $4.00 / $18.00 · tier over 200,000 tokens | 1,048,576 |
| 36 | GPT-5.6 Terra OpenAI | $1.24 | $4.00 / $18.00 · tier over 272,000 tokens | 1,050,000 |
| 38 | Claude Opus 5.5 Anthropic | $1.24 | $4.00 / $20.00 | 1,000,000 |
| 39 | Claude Opus 4.6 Anthropic | $1.55 | $5.00 / $25.00 | 1,000,000 |
| 39 | Claude Opus 4.7 Anthropic | $1.55 | $5.00 / $25.00 | 1,000,000 |
| 39 | Claude Opus 4.8 Anthropic | $1.55 | $5.00 / $25.00 | 1,000,000 |
| 39 | Claude Opus 5 Anthropic | $1.55 | $5.00 / $25.00 | 1,000,000 |
| 43 | ChatGPT (chat-latest) OpenAI | $1.56 | $5.00 / $30.00 | 400,000 |
| 44 | GPT-5.6 Sol OpenAI promo | $2.46 | $8.00 / $30.00 · tier over 272,000 tokens | 1,050,000 |
| 45 | Claude Fable 5 Anthropic | $3.10 | $10.00 / $50.00 | 1,000,000 |
| 45 | Claude Fable 5.1 Anthropic | $3.10 | $10.00 / $50.00 | 1,000,000 |
| 45 | Claude Mythos 5 Anthropic | $3.10 | $10.00 / $50.00 | 1,000,000 |
| 45 | Claude Mythos 5.1 Anthropic | $3.10 | $10.00 / $50.00 | 1,000,000 |
| 49 | GPT-5.6 Cyber OpenAI | $3.90 | $12.50 / $75.00 | 400,000 |
| 50 | GPT-6 Astra OpenAI | $6.15 | $20.00 / $75.00 · tier over 272,000 tokens | 1,050,000 |
Promotional prices: Gemini 3.6 Flash (until December 31, 2026), Gemini 3.7 Flash (until December 31, 2026), Gemini 3.8 Flash (until December 31, 2026), GPT-5.6 Sol (at least through November 21, 2026). The cost above uses the promotional price; what each maker publishes for after the promotion, if anything, is on the model's page.
The Alibaba rows use its Singapore (International) prices. For all of them the base input price in Beijing and Global is lower; each model page shows both.
DeepSeek V4.1 Flash is ranked at its list (peak) price; off-peak, 133 of the 168 hours of the week, the same request costs $0.0462. DeepSeek V4 Pro is ranked at its list (peak) price; off-peak, 133 of the 168 hours of the week, the same request costs $0.202.
Models that cannot take a 300,000-token prompt
Context window under 302,000 tokens: Claude Haiku 4.5 (200,000), Claude Opus 4.5 (200,000), Gemini Robotics ER 2 Preview (131,072), Gemini Robotics ER 2 Streaming Preview (131,072), Grok Build 0.1 (256,000), Mistral Medium 3.5 (256,000), Mistral Large 3 (256,000), Mistral Small 4 (256,000), Codestral (128,000), Ministral 3 3B (256,000), Ministral 3 8B (256,000), Ministral 3 14B (256,000), Command A (256,000), Command R7B (128,000), Command R (08-2024) (128,000), Command R+ (08-2024) (128,000).
No context window recorded on this site, so not ranked: GPT-Rosalind.
All current prices, with cached input and sources: LLM API pricing. Your own volume: cost calculator. How every figure is checked: methodology.