#tHow Many Tokens?

← All models

o3-mini: token counter & pricing

OpenAI · exact (uses official tokenizer) · pricing as of 2026-05-31.

Provider
OpenAI
API model ID
o3-mini
Context window
200,000 tokens
Input price
$1.10 per 1M tokens
Output price
$4.40 per 1M tokens
Tokenizer accuracy
exact (uses official tokenizer)
Pricing as of
2026-05-31

Open the counter to count tokens for o3-mini in real time.

What is o3-mini?

o3-mini is the smaller, cheaper member of OpenAI's o-series reasoning models. Same approach as o3 (extended internal reasoning before responding) at roughly half the price: $1.10 input / $4.40 output per 1M tokens.

For most reasoning workloads where you don't need o3's full capacity, o3-mini is the rational default.

How tokens are counted here

OpenAI's o200k_base tokenizer. Browser-side via js-tiktoken. Exact.

But: o-series produces a lot of "reasoning tokens", internal chain-of-thought tokens that count toward your output bill but don't appear in the final response. A 200-token reply might consume 1,500-3,000 output tokens of internal reasoning. The calculator above shows the visible output count; actual billed output is typically 5-15× higher.

Pricing notes

$1.10 input / $4.40 output per 1M. Cached input $0.55/M.

For 1,000 input + 200 output visible tokens (real billed output likely 1k-3k tokens with reasoning):

200k context window, same as the rest of the o-series.

When to use o3-mini

When not to use it:

Common questions

o3-mini vs o4-mini?

Both priced at $1.10/$4.40. o4-mini is the newer generation with broader reasoning improvements; o3-mini stays on the API for production workloads pinned to its behavior profile. For new work, default to o4-mini.

o3-mini vs GPT-5.2 for reasoning?

GPT-5.2 ($1.75/$14) doesn't generate the invisible reasoning tokens that inflate o-series costs. On many tasks GPT-5.2 is the cheaper effective choice; on hardest reasoning o3-mini still wins. Run your own evals.

Caching savings?

$0.55/M cached input is 50% off standard. Lower discount than the GPT-5 family (typically 90% off), this is structurally the case for o-series caching today.

Compare o3-mini to other models