Fastest, most cost-efficient version of GPT-5. Great for summarization and classification tasks with reasoning token support. Features a 400K context window and 128K max output.
GPT-5 Nano is OpenAI's smallest, fastest, and most cost-efficient GPT-5 variant, optimized for well-defined tasks where speed-to-cost ratio matters more than ceiling intelligence. Launched alongside the GPT-5 family on August 7, 2025, it shares the same 400K context window and 128K max output tokens as its larger siblings — at roughly one-tenth the cost of GPT-5 Mini and one-fiftieth the cost of GPT-5. On TheRouter, openai/gpt-5-nano is priced at $0.05 per 1M input tokens and $0.40 per 1M output tokens, making it one of the most affordable reasoning-capable models with long context support.
Despite its small size, GPT-5 Nano supports the full GPT-5 parameter suite: reasoning_effort (minimal through high), verbosity control (low/medium/high), tool calling with parallel execution, response_format for structured JSON output, and image understanding with text-only output. Its low latency profile — combined with the 90% cache-read discount ($0.005/1M cached input tokens) — makes it ideal for high-volume production workloads where quick, accurate classification or extraction is the primary goal.
| Type | Rate |
|---|---|
| Input | $0.054 / 1M tokens |
| Output | $0.432 / 1M tokens |
| Release date | 2025-08-07openai.com ↗ | verified |
| Knowledge cutoff | May 2024openrouter.ai ↗ | to verify |
| OpenAI model id | gpt-5-nano-2025-08-07platform.openai.com ↗ | to verify |
| License | OpenAI Terms of Use (proprietary, API-only) | verified |
| Benchmark | Distribution | Score | Source |
|---|---|---|---|
SWE-bench Verified As the smallest GPT-5 variant, GPT-5 Nano scores below GPT-5 (74.9%) on SWE-bench. OpenAI has not published variant-level breakdowns. GPT-5 Nano is designed for classification and extraction, not code generation. | Not publicly disclosed | openai.com ↗ | |
AIME 2025 GPT-5 family SOTA on AIME 2025: 94.6% (GPT-5, high reasoning). GPT-5 Nano variant-level score not published by OpenAI. | Not publicly disclosed | openai.com ↗ | |
MMMU (Multimodal) GPT-5 overall: 84.2%. GPT-5 Nano variant-level MMMU not separately published. | Parent: 84.2% | openai.com ↗ |
Use the global api.therouter.ai endpoint shown below for new integrations; the legacy China accelerated endpoint is retired.
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-nano",
"messages": [
{"role": "user", "content": "Summarize the key points from this input."}
]
}'GPT-5 Nano is called via the standard chat completions endpoint through TheRouter, exactly like any OpenAI-compatible model. All GPT-5 parameters are supported.
curl https://api.therouter.ai/v1/chat/completions \
-H "Authorization: Bearer $THEROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "openai/gpt-5-nano",
"messages": [
{"role": "user", "content": "Classify this text as positive, negative, or neutral: \"The service was okay but nothing special.\""}
],
"reasoning_effort": "minimal",
"response_format": { "type": "json_object" }
}'OpenAI released three GPT-5 variants — GPT-5 (flagship), GPT-5 Mini (mid-tier), and GPT-5 Nano (budget). All variants share 400K context and 128K max output. The Nano tier starts at $0.05/$0.40 per 1M tokens with support for reasoning_effort, verbosity, tool calling, and vision.
GPT-5 Nano is the smallest, cheapest GPT-5 variant ($0.05/$0.40 per 1M tokens vs GPT-5 Mini's $0.25/$2). It is optimized for high-volume classification, summarization, and extraction tasks where raw intelligence matters less than speed and cost. GPT-5 Mini handles deeper reasoning, multi-step agents, and code generation better, but costs 5× more.
Yes. GPT-5 Nano accepts image inputs (text and image modalities) and produces text-only output. It can read charts, diagrams, screenshots, and documents. It does not generate images — use gpt-image-2 for that.
Yes. GPT-5 Nano remains available on TheRouter and is still actively served by OpenAI via the API platform. However, new projects should evaluate GPT-5.4 Nano (openai/gpt-5.4-nano) as the recommended successor with improved lightweight reasoning capabilities.
GPT-5 Nano enjoys the same 90% cache-read discount as the rest of the GPT-5 family — cached input tokens cost $0.005 per 1M instead of $0.05. This makes iterative prompting patterns (e.g., few-shot classification with shared system prompts) extremely economical for production workloads.
| source | URL | retrieved | |
|---|---|---|---|
| Release date | openai.com ↗ | 2026-05-29 | verified |
| Knowledge cutoff | openrouter.ai ↗ | 2026-05-29 | to verify |
| OpenAI model id | platform.openai.com ↗ | 2026-05-29 | to verify |
| License | — | — | verified |
| SWE-bench Verified | openai.com ↗ | 2026-05-29 | to verify |
| AIME 2025 | openai.com ↗ | 2026-05-29 | to verify |
| MMMU (Multimodal) | openai.com ↗ | 2026-05-29 | to verify |
| GPT-5 family launch — GPT-5, Mini, and Nano | openai.com ↗ | 2026-05-29 | verified |
| How does GPT-5 Nano compare to GPT-5 Mini? | openai.com ↗ | 2026-05-29 | to verify |
| Does GPT-5 Nano support vision? | platform.openai.com ↗ | 2026-05-29 | to verify |
| Is GPT-5 Nano still available after GPT-5.4 Nano was released? | platform.openai.com ↗ | 2026-05-29 | to verify |
| What are the cache-read pricing benefits? | platform.openai.com ↗ | 2026-05-29 | to verify |