Know what an LLM actually costs.
Prices span 12,000× - $0.019 per million input tokens on Mistral Nemo, $150 on o1-pro. And the figure on a pricing page is often not what you pay. This tracks every model, each host that serves it, and the per-request fees most comparisons leave out.
- Models tracked
- 340
- Providers
- 52
- Gateways
- 117
- With a free tier
- 22
Labs training the models
Hosts and routers
Cheapest models right now
Ranked by blended cost - input tokens weighted 75%, output 25%, the ratio most production workloads actually hit.
| # | Model | Blended / 1M | Input / 1M | Output / 1M | Context | Capabilities |
|---|---|---|---|---|---|---|
| 1 | Mistral Nemo Mistral AI | $0.022 | $0.019 | $0.030 | 131K | ToolsOpen weights |
| 2 | Ling 3.0 Flash inclusionAI | $0.032 | $0.021 | $0.063 | 262K | ReasoningToolsOpen weights |
| 3 | Granite 4.0 Micro IBM Granite | $0.041 | $0.017 | $0.112 | 131K | Open weights |
| 4 | Llama 3 8B Lunaris Sao10K | $0.042 | $0.040 | $0.050 | 8K | Open weights |
| 5 | Qwen3.7 Flash Qwen | $0.055 | $0.030 | $0.130 | 1M | ReasoningToolsVision |
| 6 | gpt-oss-20b OpenAI | $0.055 | $0.030 | $0.130 | 131K | ReasoningToolsOpen weights |
| 7 | Mistral Small 3 Mistral AI | $0.057 | $0.050 | $0.080 | 33K | Open weights |
| 8 | Llama 3.1 8B Instruct Meta Llama | $0.057 | $0.050 | $0.080 | 131K | ToolsOpen weights |
| 9 | Schematron V2 Turbo Inference.net | $0.060 | $0.030 | $0.150 | 128K | Open weights |
| 10 | MythoMax 13B MythoMax 13B | $0.060 | $0.060 | $0.060 | 8K | Open weights |
What the expensive end buys
Frontier pricing, and the context window and capability it actually gets you.
| Model | Blended / 1M | Input / 1M | Output / 1M | Context | Capabilities |
|---|---|---|---|---|---|
o1-pro OpenAI | $263 | $150 | $600 | 200K | ReasoningVision |
GPT-5.4 Pro OpenAI | $67.50 | $30.00 | $180 | 1.1M | ReasoningToolsVision |
GPT-5.5 Pro OpenAI | $67.50 | $30.00 | $180 | 1.1M | ReasoningToolsVision |
GPT-5.2 Pro OpenAI | $57.75 | $21.00 | $168 | 400K | ReasoningToolsVision |
GPT-5 Pro OpenAI | $41.25 | $15.00 | $120 | 400K | ReasoningToolsVision |
One API key, every model
Gateways sit between you and the labs. OpenRouter is the best known - it is not the only one, and it is not always the cheapest.
AI/ML API
US
Single subscription-style endpoint fronting 300+ text, image and audio models.
Free tierEden AI
FR
One API across many AI vendors, spanning language, vision, speech and OCR.
Free tierOpenRouter
US
The reference model router: one OpenAI-compatible endpoint in front of 400+ models and 100+ hosts, with automatic price and uptime failover.
Free tierRequesty
NL
Routing layer positioned as a direct OpenRouter alternative, emphasising caching and spend controls.
Free tierVercel AI Gateway
US
One endpoint and one bill across providers, wired into the Vercel AI SDK with automatic failover.
Free tierCommon questions
Which LLM is the cheapest per million tokens?
Mistral: Mistral Nemo is currently the cheapest paid model in this directory at $0.019 per million input tokens and $0.030 per million output tokens.
How much cheaper is the cheapest model than the most expensive?
About 12,000× on a blended basis. The blend weights input tokens at 75% and output tokens at 25%, which reflects how most production workloads actually read and write.
What is an LLM gateway, and how does OpenRouter fit in?
A gateway sits between your application and the labs, exposing many models behind one API key and billing relationship. OpenRouter is the best-known example; LiteLLM, Portkey, Requesty, Vercel AI Gateway and Together AI solve overlapping parts of the same problem in different ways.
Are these prices official?
Prices are read from the live OpenRouter catalogue, which mirrors what each provider charges, and are normalised to US dollars per million tokens. Each model page links to the provider's own pricing page so you can confirm before committing spend.