What AI models cost.
Checked every morning.
LLM Price Index gathers public API price listings each morning, alongside curated maker prices. Look up a tracked model to compare listed prices, see which prices changed overnight, and find what you can run for free. Each comparison links to source evidence where available; catalog fallbacks and unknown verification dates are labeled.
Provider free tiers and zero-price models, compared on the allowance, the tightest limit, what it takes to start, and how your data is treated. Every number quoted from the provider's own page.
Compare every free program → Cheaper than the maker0 listings at least 15% below the maker's priceThe same model, sold cheaper through another provider. Listed token price only; not a claim about speed, terms or quality.
See the below-direct list → Every price755 models, with history back to July 2023Each listing's current rate, every observed repricing, the cheapest route today, and a free route where one exists.
Open the model catalog →Recent observed moves
At least 10% change in input or output price between two daily snapshots.
Find a model
USD per 1M standard text tokens. Confirm provider terms before purchase.
Comparison scope: provider prices stay paired. Blended combines each provider's input and output prices into one consistent comparison number; Input + Output shows both prices. No performance ranking is implied.
Free now & below direct
Free access is classified by offer type. Below-direct means listed token price only, not better performance or identical service terms.
Free access monitor
Zero-price models, provider free tiers, and limited offers, classified separately.
Compare every verified program →Biggest stated free allowances today. Rows link to the program's page.
| Program | Free allowance | To start |
|---|---|---|
| AIHubMixAIHubMix free model catalog | 1M tokens / dayshared pool | One-time $1 deposit |
| OpenAIFree daily tokens for opted-in data sharing | 1M tokens / dayshared pool | One-time $5 deposit |
| LLM7.ioLLM7 free-token tier | 1M tokens / day | Signup, no card |
| GroqGroq free-tier models | 200K tokens / dayper model | Signup, no payment |
| Aion LabsAion Labs API free tier | 20K tokens / day | Signup, no payment |
| IBMwatsonx.ai foundation models (playground) | 300K tokens / month | Card on file |
Below direct
Current inference-provider listings at least 15% below the curated first-party rate.
No qualifying comparison with a currently verified maker price is available in this snapshot. Missing or overdue evidence is not proof that no cheaper route exists.
Lowest recorded prices
54 current listings sit at the lowest price ever recorded for them. 5 set that low in the last 30 days.
- OpenAI: GPT-5.6 Sol Pro$4.00-50.0% · set Oct 5, 2026
- Tencent: Hy3$0.14-37.5% · set Sep 17, 2026
- Qwen: Qwen3.5-35B-A3B$0.25-31.7% · set Oct 9, 2026
- Qwen: Qwen3.5-122B-A10B$0.72-12.5% · set Sep 17, 2026
- NVIDIA: Nemotron 3 Ultra$0.93-11.9% · set Oct 2, 2026
Source: OpenRouter's public model listing, the only source with multi-year history; other columns on this site bank forward history from June 2026. Free variants and router pseudo-models are left out. Every listing's history
Source health
755 model keys tracked, 411 comparable across two or more named sources, 8 sources reporting today. “Comparable” does not mean the endpoints, terms, speed, or quality are identical. Green is fetched live; amber is retained last-good data; gray predates health tracking or is unavailable.
What these numbers do and do not mean
LLMPI compares base standard text-token listings. Blended uses the same fixed, input-heavy mix for every provider so they can be screened with one number; it is not a prediction of your exact bill. LLMPI does not normalize caching, batch discounts, long-context tiers, quantization, regions, privacy, latency, reliability, or model quality. Direct API rates are curated from first-party pages and display their own verification date instead of inheriting the newest automated refresh date.