Curiosity · AI Model
GPT-4o
GPT-4o ("omni") is OpenAI's May 2024 multimodal model — the first GPT where text, image, and audio share one neural network instead of being bolted together. It made real-time voice affordable, cut GPT-4-class pricing roughly in half, and became the default ChatGPT model for most of 2024.
Model specs
- Vendor
- OpenAI
- Family
- GPT-4
- Released
- 2024-05
- Context window
- 128,000 tokens
- Modalities
- text, vision, audio, code
- Input price
- $2.5/M tok
- Output price
- $10/M tok
- Pricing as of
- 2026-04-20
Strengths
- First true omnimodal model — audio emotion, laughter, and tone preserved
- Roughly 2x faster and 50% cheaper than GPT-4 Turbo at launch
- Strong multilingual performance, especially non-Latin scripts
- Wide ecosystem support — Azure, API, ChatGPT all shipped day-one
Limitations
- Overtaken on reasoning by o1/o3 and on coding by GPT-4.1 and GPT-5
- Native audio output gated behind Realtime API, not standard Chat Completions
- Knowledge cutoff is October 2023 — stale for current-events tasks
Use cases
- ChatGPT default consumer experience (2024-2025)
- Voice interfaces and real-time agents via Realtime API
- Image understanding — charts, screenshots, handwriting
- Cost-sensitive general-purpose chat at GPT-4 quality
Benchmarks
| Benchmark | Score | As of |
|---|---|---|
| MMLU | ≈88.7% | 2024-05 |
| HumanEval | ≈90.2% | 2024-05 |
| MGSM (multilingual math) | ≈90.5% | 2024-05 |
Frequently asked questions
What is GPT-4o?
GPT-4o ("omni") is OpenAI's multimodal model released in May 2024. It is the first GPT model trained end-to-end across text, vision, and audio, meaning one network handles all three modalities with ~320ms voice latency.
What is the context window of GPT-4o?
GPT-4o has a 128,000-token context window with up to 16,384 tokens of output per response.
How much does GPT-4o cost?
As of April 2026, GPT-4o is priced at roughly USD 2.50 per million input tokens and USD 10 per million output tokens on the OpenAI API, with cached input tokens discounted further.
Is GPT-4o still the best OpenAI model?
No. GPT-4o was flagship in 2024 but has been surpassed by GPT-4.1 for general tasks, by the o-series (o1, o3) for reasoning, and by GPT-5 tiers for 2026 work. It remains a solid cost-effective choice for everyday multimodal chat.
Sources
- OpenAI — Hello GPT-4o — accessed 2026-04-20
- OpenAI — Pricing — accessed 2026-04-20