GPT-5 Mini API Cost: $0.25/$2.00 Per Million Tokens

GPT-5 Mini costs $0.25/1M input and $2.00/1M output tokens — a 272K context budget model that delivers GPT-5-class reasoning at 90% less than GPT-5.4. Here's exactly what it costs per request, per 1K requests, and per month across 5 real workloads. Note: GPT-5 Mini shuts down Dec 11, 2026; consider GPT-5.6 Terra for new projects.

For comparison, GPT-4o mini (the model GPT-5 Mini replaces) was $0.15/$0.60. GPT-5 Mini costs 67% more on input but delivers significantly better reasoning and instruction-following. The output price jump from $0.60 to $2.00 is steeper, but the quality improvement often means fewer retries and shorter prompts — netting out cheaper in practice.

Cost Per Request

Here's what a single API call costs with GPT-5 Mini, based on real-world request sizes:

The sweet spot for GPT-5 Mini is 500-1,000 token requests — typical chatbot and assistant interactions. At $0.001 per request, you get 1,000 API calls for roughly $1.

Cost Per 1K Requests

Scaling up to 1,000 requests gives a clearer picture of real-world costs:

At 10,000 requests per day with medium chat responses, GPT-5 Mini costs $33.75/month. That's production-grade AI for less than a Netflix subscription.

Workload Cost Breakdowns

1. Customer Support Chatbot

Typical setup: 500 input tokens (system prompt + history), 200 output tokens per response, 1,000 conversations/day.

Compare: GPT-5.4 at $2.50/$15 would cost $210/month for the same workload. GPT-5 Mini saves 92%.

2. Code Generation Assistant

Typical setup: 1,000 input tokens (code context + instructions), 800 output tokens, 500 requests/day.

Compare: DeepSeek V4 Pro at $0.44/$0.87 would cost $13.05/month — cheaper but with different reasoning characteristics. GPT-5 Mini offers better instruction-following for complex code tasks.

3. RAG Pipeline

Typical setup: 2,000 input tokens (retrieved context + question), 300 output tokens, 2,000 queries/day.

For RAG workloads where input context dominates, the input token price matters most. GPT-5 Mini at $0.25/1M input is competitive with Gemini Flash ($0.10) when you factor in quality.

4. Content Writing / Summarization

Typical setup: 2,000 input tokens (source material), 2,000 output tokens (generated content), 200 requests/day.

5. High-Volume Classification

Typical setup: 200 input tokens, 50 output tokens, 10,000 requests/day.

For high-volume, short-output tasks, GPT-5 Mini is extremely competitive. Qwen 3.7 Flash at $0.03/$0.13 would cost $5.10/month — 72% cheaper — but GPT-5 Mini handles more nuanced classification.

GPT-5 Mini vs Alternatives

How does GPT-5 Mini stack up against other budget models?

Key insight: GPT-5 Mini is not the cheapest budget model. Qwen 3.7 Flash ($0.03/$0.13), DeepSeek V4 Flash ($0.14/$0.28), and GPT-5 nano ($0.05/$0.40) are all cheaper. GPT-5 Mini's advantage is quality per dollar — it delivers GPT-5-class reasoning at budget pricing. For tasks where accuracy matters more than raw cost, GPT-5 Mini is the sweet spot. Note: GPT-5 Mini shuts down Dec 11, 2026 — consider GPT-5.6 Terra for new projects.

When to Use GPT-5 Mini (and When Not To)

Use GPT-5 Mini when:

• You need GPT-5 quality at budget prices — reasoning, instruction-following, and nuance matter
• Your workload is chatbot, assistant, or RAG — the 272K context and strong reasoning shine here
• You're upgrading from GPT-4o mini and want better quality without jumping to GPT-5 pricing
• You need OpenAI ecosystem compatibility — same API, same SDK, same tools

⚠️ Sunset warning: GPT-5 Mini shuts down Dec 11, 2026. For new projects, start with GPT-5.6 Terra (the replacement) or GPT-5 nano (cheapest OpenAI option).

Use something cheaper when:

• High-volume classification — Qwen 3.7 Flash at $0.03/$0.13 is 88% cheaper on input
• Simple chatbots — DeepSeek V4 Flash at $0.14/$0.28 delivers comparable quality at 44% less
• Long-context analysis on a budget — Gemini 2.5 Flash-Lite offers 1M context at $0.10/1M input
• Cheapest OpenAI model — GPT-5 nano at $0.05/$0.40 is 80% cheaper on input

Batch API: Cut Costs by 50%

OpenAI's Batch API processes requests asynchronously within 24 hours at 50% off standard pricing. For GPT-5 Mini, that means:

• Batch input: $0.125/1M tokens
• Batch output: $1.00/1M tokens

If your workload isn't time-sensitive — overnight processing, data enrichment, bulk classification — the Batch API makes GPT-5 Mini competitive with Gemini Flash on price while maintaining GPT-5 quality.

— See if you're overpaying for AI APIs

Is GPT-5 Mini Being Deprecated?

Yes — GPT-5 Mini is scheduled for shutdown on December 11, 2026. OpenAI's deprecation schedule confirms it will be replaced by GPT-5.6 Terra. If you're building new applications, consider starting with one of the replacements instead:

• GPT-5.6 Terra — the direct replacement for GPT-5 Mini. Similar budget tier, improved capabilities.
• GPT-5.6 Luna ($0.20/$1.20, 1.05M context) — a step up in capability with massive context window. 4x the input price but far more capable for complex tasks.
• GPT-5 nano ($0.05/$0.40, 128K) — if you need the cheapest OpenAI option, nano is 80% cheaper on input and 83% cheaper on output.

Existing GPT-5 Mini integrations will continue working until December 11. Plan your migration by October to avoid last-minute scrambles.

The Bottom Line

GPT-5 Mini at $0.25/$2.00 occupies a unique position in the 2026 budget model landscape. It's not the cheapest option — but it's the cheapest model that delivers GPT-5-class reasoning. For developers who need quality without the GPT-5 price tag, it's the obvious choice.

At typical chatbot volumes (1K requests/day), you're looking at $15-35/month for a production-grade AI assistant. That's less than most SaaS subscriptions — and you get access to one of the most capable language models available.