Why don’t you show speed in the price table?

Price is catalog data; speed is measured in the arena when we have a replay. Don’t mix them in one number without weights — that’s what the leaderboard sliders are for.

Is OpenRouter cheaper than calling OpenAI directly?

Sometimes, via routing; sometimes not. Compare the row you will actually call.

Do you include image token pricing?

We surface catalog text token prices. Vision billing can add image units — confirm on the provider for multimodal jobs.

How often do prices update?

We revalidate catalog fetches about every hour.

LLM pricing comparison — input, output and blend

LLM pricing comparison is mostly unit conversion: everything is dollars per million tokens, but workloads are not 50/50 input/output. This page is how to use live list prices without pretending they are invoices.

Use the compare tool for two to four models’ input, cached input, and output rows. Use model detail pages for a monthly cost table at fixed volumes. Use the leaderboard blend when you only need a rank.

Cheapest models in the live catalog

Live OpenRouter pricing and Arena quality, cached about an hour. Not a static blog table.

Model	Provider	Blend $ / 1M	Quality	Context
Mistral Nemo	Mistral	0.02	—	131,072
Ling 3.0 Flash	inclusionAI	0.04	—	262,144
Llama 3 8B Lunaris	Sao10k	0.04	—	8,192
MythoMax 13B	Gryphe	0.06	—	4,096
Nex-N2-Mini	Nex Agi	0.06	—	262,144

List price vs contract price

OpenRouter and lab list prices are public. Enterprise commits, batch APIs, and reserved capacity can be lower. This site is a public-list comparison so you can argue from the same numbers; your account manager can still beat them.

Batch and cache

If 70% of tokens are a repeated system prompt plus retrieved docs, cached input pricing (when present) matters more than the headline completion price. Look at the compare table’s cached input row.

Reasoning token accounting

Thinking models may bill extra tokens you never show the user. Log usage. A “cheap” reasoning model can lose to a mid-tier non-reasoning model on real invoices.

Currency and units

We display USD per million tokens converted from OpenRouter’s per-token floats. Tiny rounding differences vs a vendor’s marketing page are normal.

Building a pricing sheet

Export Markdown from compare, add your expected input/output ratio, and multiply. For a single model, the 1/10/100M table is faster. Recheck monthly — the catalog is hourly, your sheet should not be annual.

FAQ

Why don’t you show speed in the price table?: Price is catalog data; speed is measured in the arena when we have a replay. Don’t mix them in one number without weights — that’s what the leaderboard sliders are for.
Is OpenRouter cheaper than calling OpenAI directly?: Sometimes, via routing; sometimes not. Compare the row you will actually call.
Do you include image token pricing?: We surface catalog text token prices. Vision billing can add image units — confirm on the provider for multimodal jobs.
How often do prices update?: We revalidate catalog fetches about every hour.

LLM pricing comparisonAI model pricingcost per million tokensGPT vs Claude price

Open the compare tool