OpenAI
OpenAI in 2026 ships GPT-6 Astra, Sora 2 video generation, and Codex coding agents, with API pricing from $0.20 to $30 per million tokens by model tier.
About OpenAI
What is OpenAI (2026 update)
OpenAI is the AI research and deployment company behind ChatGPT, the GPT model family, the Sora video generator, and the Codex coding agent. Beyond the consumer chatbot, OpenAI sells API access to its models by the token, along with standalone products for video generation and software development, and by September 2026 its lineup spans several actively priced model generations at once.
What is new in 2026
OpenAI’s newest model, GPT-6 Astra, began rolling out on September 3, 2026. OpenAI describes Astra as its most capable model broadly deployed to date and its first model to reach the “Critical” cybersecurity capability threshold under the company’s own Preparedness Framework, a classification OpenAI uses to flag models needing extra safety precautions. Earlier in the year, the GPT-5.6 family reached general availability on July 9, 2026, with a 1.05 million token context window across all three tiers (Sol, Terra, and Luna), and GPT-5.5 launched April 24, 2026. On the product side, Sora 2 shipped as OpenAI’s flagship video and audio generation model, improving realism, physics simulation, audio, and creative control, and generating short videos from 5 to 60 seconds long for ChatGPT Plus and Pro subscribers. A Sora Android app followed, reaching #1 on the Play Store on launch day with users generating over a million videos in its first 24 hours. Codex, OpenAI’s coding agent, gained a Slack integration, a Codex SDK, and enterprise controls, and was used internally to build the Sora Android app in 28 days.
Key features
- GPT-6 Astra, OpenAI’s most capable broadly-deployed model as of September 2026
- Sora 2 video and audio generation, producing 5-60 second clips from text prompts
- Codex coding agent with Slack integration, a developer SDK, and enterprise admin controls
- Multiple concurrently priced API model tiers (GPT-5, GPT-5.4, GPT-5.5, GPT-5.6 Sol/Terra/Luna) for different cost-performance tradeoffs
- Cached input pricing at a fraction of standard token rates for repeated context
- Batch API for non-realtime workloads at reduced cost
Pricing in 2026
OpenAI’s API is billed per million tokens, with older and newer model generations priced side by side.
| Plan | Price | What you get |
|---|---|---|
| GPT-5 API | From $0.625 input / $5.00 output per 1M tokens | Entry-level pricing tier of the GPT-5 line |
| GPT-5.4 API | From $2.50 input / $15.00 output per 1M tokens | Released March 5, 2026, mid-tier reasoning and quality |
| GPT-5.5 API | $5 input / $30 output per 1M tokens | Released April 24, 2026, OpenAI’s flagship model at the time |
| GPT-5.6 Luna API | $0.20 input / $1.20 output per 1M tokens | Cheapest of the GPT-5.6 tier after a July 30, 2026 price cut, 1.05M token context |
| GPT-5.6 Terra API | $2 input / $12 output per 1M tokens | Mid-tier GPT-5.6 model, 1.05M token context |
| GPT-5.6 Sol API | $5 input / $30 output per 1M tokens | Top-tier GPT-5.6 model, 1.05M token context |
| Cached input / Batch API | 10% of standard rate / roughly 50% off | Discounts for repeated context (caching) and non-realtime batch jobs |
GPT-6 Astra pricing was not confirmed in available sources at time of writing, so check the official OpenAI API pricing page for current Astra rates.
Who should use it
Developers building cost-sensitive applications benefit from GPT-5.6 Luna’s low per-token price with a large context window. Teams needing top reasoning quality or planning to use GPT-6 Astra should budget for premium per-token rates. Video creators and marketers get direct value from Sora 2’s short-form video generation, while engineering teams automating pull requests and multi-step coding tasks are the core audience for Codex.
Limitations and alternatives
Running multiple GPT-5.x generations concurrently makes OpenAI’s pricing page harder to read than a single flat rate, and per-model prices have changed several times in 2026, so always verify current rates before budgeting a project. GPT-6 Astra’s “Critical” cybersecurity classification signals OpenAI itself expects added safeguards and possibly restricted access for some use cases; check the official site for availability. For API alternatives with different cost curves, Google’s Gemini Flash/Pro line and Anthropic’s Claude models are the most common points of comparison. For AI video generation specifically, Google’s Veo and Runway are frequently benchmarked against Sora 2.
Frequently asked questions
What is OpenAI's latest model in September 2026?
GPT-6 Astra, which began rolling out on September 3, 2026 and is described by OpenAI as its most capable broadly-deployed model to date.
How much does the OpenAI API cost in 2026?
It depends on the model. Prices range from about $0.20 input / $1.20 output per million tokens for GPT-5.6 Luna up to $5 input / $30 output per million tokens for GPT-5.5 and GPT-5.6 Sol, check the official pricing page for current Astra rates.
What is Sora 2?
Sora 2 is OpenAI's flagship video and audio generation model, producing 5 to 60 second video clips from text prompts, available to ChatGPT Plus and Pro subscribers and through a dedicated Android app.
What is Codex used for?
Codex is OpenAI's coding agent that can complete multi-step software tasks; in 2026 it gained a Slack integration, a developer SDK, and enterprise admin controls, and was used internally to build the Sora Android app in 28 days.
Information last verified: September 2026.
Related Tools
