Every model with a :batch id on OpenRouter, priced for a job you set below. Tiers are set by batch output price: Cheap under $3 per million output tokens, Mid $3–20, Expensive $20 and up. Data: OR-Batch-Inference-Models.
Each tile shows the cheapest and dearest model in that sub-band for the job above. The models are grouped by output price, so on input-heavy jobs the cost ranges of neighbouring bands can overlap.
Batch prices with reasoning tokens excluded. Log scale. Hover a bar for prices.
US$ per million tokens, log–log. Dashed lines mark the tier boundaries at $3 and $20 output. Models above the dotted line charge more than 5× input for output. Ringed points have input priced in a different tier from their output.
Stable = the id always resolves to one model version. OpenRouter's -latest aliases have no batch ids. GPT-4o, GPT-4o mini and GPT-4 Turbo are marked No: their slugs carry no date, so OpenAI can repoint them. Hover the cell for the reason. Columns are documented in the dataset's data dictionary.
Prices change often: re-check on openrouter.ai before relying on them. Rubric and method: OpenRouter-Batch-Notes, entry 07.