TheRouter.aiTheRouter.ai
Back to Models

GPT-5 Mini

openaiopenai/gpt-5-mini

A faster, cost-efficient version of GPT-5 for well-defined tasks. Features reasoning token support with a 400K context window and 128K max output at a fraction of the cost.

Introducing GPT-5 for developers (GPT-5 Mini announcement) ↗OpenAI GPT-5 Mini developer docs ↗

GPT-5 Mini is a faster, cost-efficient variant of GPT-5 aimed at well-defined reasoning tasks that don't require the full depth of GPT-5 or GPT-5 Pro. Launched alongside the GPT-5 family on August 7, 2025, it offers the same 400K context window and 128K max output tokens as its larger sibling, with support for the full GPT-5 parameter suite including reasoning_effort, verbosity, tool calling, structured output, and vision. On TheRouter, openai/gpt-5-mini is priced at $0.25 per 1M input tokens and $2 per 1M output tokens — roughly one-fifth the cost of GPT-5 — making it a strong candidate for high-volume production workloads.

GPT-5 Mini is not just a downsized GPT-5 — it is a separate model trained cost-efficiently for well-defined, structured tasks. It supports the same reasoning_effort parameter (minimal through high), verbosity control (low/medium/high), custom_tools with grammar constraints, parallel tool calls, and response_format for structured JSON output. Combined with its low latency profile, GPT-5 Mini is well-suited for classification, summarization, data extraction, tool orchestration, and any workload where speed-to-quality ratio matters more than ceiling intelligence.

Best for
  • • High-volume classification and data extraction — fast structured outputs at $2/M output tokens
  • • Summarization and rewriting — long-context (400K) at a fraction of GPT-5 cost
  • • Tool orchestration and multi-step agents — full tool-calling with parallel execution at low latency
  • • Vision-enabled tasks with text output — read charts, diagrams, and screenshots efficiently
  • • Caching-friendly workloads — 90% input cache discount ($0.025/1M cached tokens) makes iterative prompting economical
Reach for something else if
  • • Deep reasoning or frontier math — use GPT-5 or GPT-5 Pro for AIME-level problems and complex agent chains
  • • Image or audio generation — text-only output via API; use gpt-image-2 or gpt-realtime-2 for media generation
  • • Open-weight fine-tuning — proprietary model, API-only access; use GPT-4o-mini or a local LLM for fine-tuning scenarios
  • • Very latency-sensitive real-time voice — use gpt-realtime-1.5 or gpt-realtime-2 for sub-300ms voice interactions
Context Length
400K
Max Output
128K
Input Priceper 1M tokens
$0.270/ 1M tokens
Output Priceper 1M tokens
$2.16/ 1M tokens

Modalities

textimage→text

Capabilities

Vision

Pricing Breakdown

TypeRate
Input$0.270 / 1M tokens
Output$2.16 / 1M tokens

Supported Parameters

temperaturemax_tokenstop_ptoolstool_choiceresponse_formatreasoningstop

Specifications

Release date2025-08-07openai.com ↗verified
Knowledge cutoffMay 2024artificialanalysis.ai ↗to verify
OpenAI model idgpt-5-mini-2025-08-07platform.openai.com ↗verified
LicenseOpenAI Terms of Use (proprietary, API-only)verified

Benchmarks

BenchmarkDistributionScoreSource
SWE-bench Verified
As a sub-size of the GPT-5 family, GPT-5 Mini scores below GPT-5 (74.9%). Independent evaluation by SWE-bench team found GPT-5 Mini competitive on a minimal agent configuration. Full results per reasoning effort level were released and subsequently retracted per OpenAI's request; no official variant-level breakdown is published.
Not publicly disclosed per model variantreddit.com / SWE-bench team AMA ↗
AIME 2025
GPT-5 family SOTA AIME 2025: 94.6% (GPT-5, high reasoning). GPT-5 Mini variant-level score not published by OpenAI.
Not publicly disclosedopenai.com ↗
MMMU (Multimodal)
GPT-5 overall: 84.2%. GPT-5 Mini variant-level MMMU not separately published.
Parent: 84.2%openai.com ↗

API Usage Examples

Use the global api.therouter.ai endpoint shown below for new integrations; the legacy China accelerated endpoint is retired.

cURL
curl https://api.therouter.ai/v1/chat/completions   -H "Content-Type: application/json"   -H "Authorization: Bearer $THE_ROUTER_API_KEY"   -d '{
    "model": "openai/gpt-5-mini",
    "messages": [
      {"role": "user", "content": "Summarize the key points from this input."}
    ]
  }'

API guide

Full API reference →

Chat completion

GPT-5 Mini is called via the standard chat completions endpoint through TheRouter, exactly like any OpenAI-compatible model. All GPT-5 parameters are supported: reasoning_effort, verbosity, response_format, tools, and stop sequences.

cURL
curl https://api.therouter.ai/v1/chat/completions \
  -H "Authorization: Bearer $THEROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-5-mini",
    "messages": [
      {"role": "user", "content": "Classify the sentiment of this customer review as positive, negative, or neutral: \"The product arrived on time but the packaging was damaged.\""}
    ],
    "reasoning_effort": "low",
    "response_format": { "type": "json_object" }
  }'

More from openai

openai/gpt-5Full-size GPT-5 for deep reasoning and codingopenai/gpt-5-nanoSmallest and cheapest GPT-5 variant for basic tasksopenai/gpt-5-proExtended reasoning GPT-5 for expert-level answersopenai/gpt-4o-miniPrevious generation small model — lower cost, smaller context (128K)openai/gpt-4.1-miniPrevious generation mid-tier mini — 1M context, no reasoning

Similar models

Cross-provider sibling models
anthropic/claude-haiku-4.5
Anthropic's fast, small model — similar price, 200K context
google/gemini-2.5-flash
Google's fast reasoning model — 1M context, competitive pricing
deepseek/deepseek-v3.2
DeepSeek's cost-efficient chat model — high context, very low pricing
mistral/mistral-small
Mistral's small model — open-weight option, comparable price tier

News & changes

2025-08-07

OpenAI launches GPT-5 family: GPT-5, GPT-5 Mini, GPT-5 Nano

OpenAI released the GPT-5 family on August 7, 2025, featuring three sizes: GPT-5 ($1.25/$10), GPT-5 Mini ($0.25/$2), and GPT-5 Nano ($0.05/$0.40). All three share 400K context, 128K max output, and support reasoning_effort, verbosity control, tool calling, structured output, and vision. The launch also introduced custom_tools with grammar constraints, a minimal reasoning effort setting, and improved tool intelligence for agentic workflows.

re-authored by TheRouteropenai.com ↗

Frequently asked

How does GPT-5 Mini differ from GPT-5?

GPT-5 Mini is a separate model trained for well-defined, structured tasks at roughly one-fifth the cost of GPT-5 ($0.25/$2 vs $1.25/$10 per 1M tokens). Both share 400K context and the same parameter set, but GPT-5 Mini is optimized for speed and cost-efficiency rather than ceiling intelligence. For deep reasoning, frontier math, or complex agentic coding, GPT-5 or GPT-5 Pro are better choices.

re-authored by TheRouterartificialanalysis.ai ↗
Can GPT-5 Mini be used through the standard OpenAI SDK?

Yes. On TheRouter, GPT-5 Mini is fully compatible with the OpenAI SDK. Simply set baseURL to https://api.therouter.ai/v1 and use 'openai/gpt-5-mini' as the model identifier. All parameters including reasoning_effort, verbosity, tools, response_format, and streaming work unchanged.

Fact ledger — every claim on this page traces here
sourceURLretrieved
Release dateopenai.com ↗2026-05-28verified
Knowledge cutoffartificialanalysis.ai ↗2026-05-28to verify
OpenAI model idplatform.openai.com ↗2026-05-28verified
License——verified
SWE-bench Verifiedreddit.com / SWE-bench team AMA ↗2026-05-28to verify
AIME 2025openai.com ↗2026-05-28to verify
MMMU (Multimodal)openai.com ↗2026-05-28to verify
OpenAI launches GPT-5 family: GPT-5, GPT-5 Mini, GPT-5 Nanoopenai.com ↗2026-05-28verified
How does GPT-5 Mini differ from GPT-5?artificialanalysis.ai ↗2026-05-28to verify
Help & contact