If you call a language model from your own code, you pay per token, not per month. The cheapest API is the lowest-priced model that still does your task well. Below are standard-rate prices for short prompts, from our verified dataset, sorted by input price.
Ranked by input price (standard tier, short prompts)
| Model | Input / 1M tokens | Output / 1M tokens | Notes |
|---|---|---|---|
| OpenAI GPT-6 Luna | $0.10 | $0.50 | Prompts ≤272K tokens |
| Anthropic Claude Haiku 5.5 | $0.10 | $0.50 | Prompts ≤100K tokens; above that $0.50 / $2.50 |
| DeepSeek Flash | $0.30 | $1.20 | Peak, cache miss. Off-peak $0.15 / $0.60 |
| Google Gemini 3.5 Flash-Lite | $0.30 | $2.50 | Output includes thinking tokens |
| Google Gemini 3.8 Flash | $0.75 | $3.75 | Promo until 2026-12-31; from 2027-01-01 scheduled at $1.50 / $7.50 |
| Perplexity Sonar | $1 | $1 | Plus a request fee of $5–$12 per 1,000 requests by search context size |
| xAI Grok 4.3 | $1.25 | $2.50 | Prompts under 200K tokens |
| DeepSeek V4 Pro | $1.32 | $3.96 | Peak; off-peak $0.66 / $1.98 |
Mid-range models for comparison: OpenAI GPT-6.1 Sol, Claude Sonnet 5.5 and Gemini 4 Argon are each $2 input / $10 output. Argon’s rate is a promo until 2026-11-30, scheduled to rise to $4 / $20 from 2026-12-01.
Batch: half price if you can wait
If your jobs don’t need an instant answer, Batch APIs cut the bill. OpenAI, Anthropic and Google all list a 50% batch discount on these models. That takes GPT-6 Luna and Claude Haiku 5.5 (short prompts) to $0.05 input / $0.25 output, and Gemini 3.5 Flash-Lite to $0.15 / $1.25. xAI lists 20% off for Grok 4.3 and no batch discount for Grok 4.7. DeepSeek’s off-peak pricing works in a similar spirit: same model, lower rate at quieter hours.
Cheapest is not just the input price
- Output usually costs more than input. A chat app that writes long answers is driven by the output rate.
- Prompt length matters. OpenAI, Anthropic and xAI charge more above a context threshold.
- Caching helps repeat prompts. Cached input is far cheaper than fresh input on every vendor we track.
- Search APIs add per-request fees, as with Perplexity Sonar.
To estimate a real monthly bill, plug your token counts into our API cost calculator. Full vendor tables are on LLM API pricing. Looking for no-cost options? See free LLM APIs.
How we check prices
AI Stack Wise reads API prices only from official vendor pricing docs, records each rate exactly as printed, and marks promo rates with their end date so they drop off automatically. OpenAI, Anthropic, Google, xAI and Perplexity rates were last checked on 2026-10-19; DeepSeek on 2026-11-03. Prices are in USD and exclude tax unless stated. See our methodology. Spotted a change? Email [email protected].