o4-mini pricing and specifications

Pricing

What o4-mini costs to run.

Input

USD 1.10

per 1M tokens

Output

USD 4.40

per 1M tokens

Price detail

Input: USD 1.10 per 1M tokens
Output: USD 4.40 per 1M tokens
Cached input read: USD 0.275 per 1M tokens
Cache write: Not published

Capabilities

Documented by the provider, not measured by Router.

Model behavior

Reasoning: Supported
Streaming: Not published

Modalities

Input: Text and image
Output: Text

Tool use

Function calling: Supported
Structured output: Supported

Delivery

Prompt caching: Not published
Fast mode: Available

Providers and keys

Bring your own provider key, or use Router’s shared capacity.

Provider billing

Router does not charge for requests served with your key — your provider bills you directly. If Router falls back to its shared key, that request is billed normally.

Router supports provider keys for Google, OpenAI, Fireworks, xAI, and Baseten, with additional providers depending on your deployment. Each user can save one key per provider.

Manage provider keys API keys How provider keys work

Set up Router

One command installs the CLI and walks you through configuring Router for your coding agents.

curl -fsSL https://router.com/install.sh | sh

ramp router configure opens an interactive agent picker. To use your own provider key, see Providers and keys.

Supported parameters

reasoning_effort: Low → high
max_output_tokens: Up to 100,000
stream: Not published
tools: Function calling
Structured output: Supported

Set up your coding agent

o4-mini is our latest small o-series model. It's optimized for fast, effective reasoning with exceptionally efficient performance in coding and visual tasks. It's succeeded by GPT-5 Mini. The published context window is 200,000 tokens, with up to 100,000 output tokens.

Creator: OpenAI

Released Apr 16, 2025. Reasoning is listed at low → high effort. Function calling is supported. Structured output is supported. Cached-read and cache-write rates are listed above where the provider publishes them.

What to consider

Reasoning cannot be disabled

The listed effort range is low → high, with no “none” option. Allow for reasoning in your output token budget.

Context and output limits

The context window is 200K tokens, with output capped at 100K. Plan for both your input and the response.

Cached reads are billed separately

Repeated prompt prefixes read from cache at USD 0.275 per 1M tokens, against USD 1.10 uncached. No separate cache-write rates are listed.

Evaluate provider performance

Check latency and throughput on your own workload. Prices and token limits do not measure serving performance.

When to use o4-mini

Recommendations are limited to what the catalog and provider documentation support.

Best for

Long documents

A 200K-token context window at USD 1.10 per 1M input tokens. Include the response when planning your token budget.

Agentic tool use

Function calling and structured output are listed, alongside reasoning at low → high effort. Evaluate tool behavior on your own tasks.

Structured responses

Structured output is listed for responses that need a defined format. Validate results against your application’s requirements.

Consider alternatives when

You need lower latency

The catalog does not publish latency measurements. Test serving performance on your workload before choosing a model.

Cost is the binding constraint

GPT-5 nano lists USD 0.05 / USD 0.40 per 1M input / output tokens, with a 400K-token context window and a 128K-token output limit. Its listed capabilities match; evaluate quality on your tasks. View model

You need reasoning switched off

Claude Haiku 5.5 lists a “none” reasoning effort, a 1M-token context window and a 128K-token output limit. o4-mini has no “none” option. View model

Compare models

Cost by workload

Estimated monthly cost at published base rates. Uncached, before service tiers or negotiated terms.

Workload	Tokens in / out	Requests / mo	Est. monthly
Short chat	1K / 500	10,000	$33
Document summary	50K / 2K	5,000	$319
Long-context agent	200K / 5K	2,000	Exceeds token limit

Frequently asked questions

o4-mini is our latest small o-series model. It's optimized for fast, effective reasoning with exceptionally efficient performance in coding and visual tasks. It's succeeded by GPT-5 Mini. It is a model from OpenAI, with a 200,000-token context window.

USD 1.10 per million input tokens and USD 4.40 per million output tokens at the published base rates. Caching, service tiers, and negotiated terms may change the final cost.

200,000 tokens, with a maximum output of 100,000 tokens.

Function calling is supported. Structured output is supported. Reasoning efforts: low → high.

Text and image input, with text output. Check the provider documentation for input limits and supported formats.

Check GET /v1/models for the callable model IDs and routes available to your API key.

Provenance

Model identity, capabilities, limits, and base prices come from the shared Router catalog. This page uses the same data as the model directory, last updated Oct 26, 2026.

Router model catalog Model documentation OpenAI release source Artificial Analysis

First recorded in catalog

Apr 9, 2025

Price change history

Not recorded

Data handling

Provider-side ZDR eligibility for o4-mini is not confirmed in the published provider schedule. Consult the current provider terms.

* Provider-side ZDR is separate from Router content recording, which is on by default. Turning recording off in Settings only affects future content after the change propagates; existing archives are not deleted and request metadata may remain. BYOK follows your direct provider agreement.

Provider retention policies Data retention and processing Content recording settings

o4-mini