The cheapest LLMs with a 1M token context window
A million tokens of context used to mean a premium price tag. Not anymore. 164 models on OpenRouter now offer 1M+ tokens of context, and the cheapest costs 1.2 cents per 100K input tokens.
The cheapest 1M-context models, ranked by input price
Prices are per 1M tokens, pulled live from the OpenRouter catalog.
| Model | Context window | Input /1M | Output /1M |
|---|---|---|---|
| DeepSeek V4 Flash Latest | 1,310,720 | $0.012 | $1.25 |
| DeepSeek V4 Flash 0731 | 1,310,720 | $0.018 | $0.32 |
| DeepSeek Flash Latest | 1,048,576 | $0.02 | $0.6 |
| GLM Flash Latest | 1,310,720 | $0.02 | $0.3 |
| Qwen3.7 Flash | 1,000,000 | $0.03 | $0.13 |
| Qwen3.8 27B | 1,000,000 | $0.0307 | $4.4 |
| GPT-6 Luna Pro (batch) | 1,050,000 | $0.05 | $0.25 |
| GPT-6 Luna (batch) | 1,050,000 | $0.05 | $0.25 |
| Gemini 2.5 Flash Lite (batch) | 1,048,576 | $0.05 | $0.2 |
| GPT-4.1 Nano (batch) | 1,047,576 | $0.05 | $0.2 |
What you can actually fit in 1M tokens
A million tokens is roughly 750,000 words — about seven average novels, a mid-sized codebase, or six months of a busy team’s chat logs. If your workflow involves whole-document analysis, long transcripts, or big code contexts, that window removes the chunking-and-stitching problem entirely.
What it costs to run heavy daily use
Here’s a concrete scenario: 500K input tokens and 50K output tokens every day, billed over 30 days.
| Model | Cost per day | Cost per month |
|---|---|---|
| DeepSeek V4 Flash Latest | $0.07 | $2.06 |
| DeepSeek V4 Flash 0731 | $0.03 | $0.75 |
| DeepSeek Flash Latest | $0.04 | $1.20 |
| GLM Flash Latest | $0.03 | $0.75 |
| Qwen3.7 Flash | $0.02 | $0.64 |
| Qwen3.8 27B | $0.24 | $7.06 |
When a premium long-context model is worth it
The cheap end of the table is fast, light models tuned for retrieval-style work. If your task is dense multi-document reasoning — comparing 40 contracts, debugging across an entire repo — quality still scales with model size. The premium end of the 1M-context market:
| Model | Input /1M | Output /1M | Context window |
|---|---|---|---|
| Claude Fable 5 | $10 | $50 | 1,000,000 |
| GPT-5.5 Pro (batch) | $15 | $90 | 1,050,000 |
| GPT-5.4 Pro (batch) | $15 | $90 | 1,050,000 |
| GPT-5.5 Pro | $30 | $180 | 1,050,000 |
| GPT-5.4 Pro | $30 | $180 | 1,050,000 |
That’s a 2,500x spread between the cheapest and priciest 1M-context model. Start cheap, and only pay flagship rates when output quality actually blocks you.
Run the numbers yourself
Use the OpenRouter pricing calculator to price your own token volumes across all 434 models, or browse related answers like GPT-5.5 Pro pricing and batch API discounts. The full list lives on the answers index.