# OpenAI released GPT Luna 5.7+ and topped key AI benchmarks

*Science. 51% certain. Thursday, 1 October 2026*

> Speculative fiction assembled from prediction-market prices. This has not happened.

**The firm also launched a dedicated Cyber model for Trusted users and labeled a model as Critical.**

OpenAI accelerated its model pipeline in late September 2026, unveiling GPT Luna 5.7+. The company also launched a dedicated Cyber model for Trusted Access users. In the same period OpenAI classified one of its models as "Critical" under its Preparedness Framework. The releases coincided with OpenAI's models attaining top rankings in both Code-WebDev and text-to-Image benchmarks.

At the end of September OpenAI's text-to-Image system ranked first among competing systems. The assessment placed OpenAI ahead of other providers in the benchmark. This result reflected the rapid development emphasized by the September releases.

In the Code Arena WebDev competition OpenAI's model secured the leading position. Anthropic's offering placed Second in the same contest. The ranking highlighted OpenAI's strength in web-development Code generation. The contest measured performance across a range of development tasks.

OpenAI's AI Agent Lab was judged the Second-best system at the close of September. The top spot was taken by another provider, leaving OpenAI just behind the leader. The ranking showed OpenAI's agents remained competitive despite not taking first place.

GPT Luna 5.7+ was released to users by September 30, meeting the company's internal schedule. A separate new model also became publicly available before October. The dedicated Cyber model was broadly released to Trusted Access users by the same deadline. The "Critical" classification applied to one of these models under the Preparedness Framework.

## Analysis

By compressing the timeline for major releases, the firm may capture a larger share of emerging applications that depend on cutting-edge language and safety features. This pace could also pressure competitors to shorten their own development cycles to remain relevant.

The text-to-Image lead may attract creators and enterprises seeking generative visual content, while the Code-WebDev advantage could deepen integration with software development workflows. Competitors who previously led in these areas now face a gap that may affect user adoption and partnership opportunities.

Classifying a model as "Critical" under the Preparedness Framework and offering a Cyber model to Trusted Access users highlights an emphasis on security and responsible deployment. The designation suggests the model carries capabilities that require heightened oversight, potentially reducing misuse. Providing a specialized Cyber model to a vetted audience may foster trust among high-risk sectors such as finance or defense, but it also raises questions about access equity.

Uncertainty remains about how other AI firms will respond to OpenAI's accelerated output and benchmark dominance. Regulatory bodies may scrutinize the rapid releases, especially the "Critical" classification, to ensure compliance with emerging safety standards. Additionally, sustaining performance leadership will depend on continued research investment and the ability to address any emerging technical or ethical challenges.

## What this was built from

- **Which company has the best Code Arena WebDev AI model end of September?**: OpenAI, 75%; settles Thursday, 1 October 2026 (polymarket). [Market](https://polymarket.com/event/which-company-has-the-best-code-arena-webdev-ai-model-end-of-september-20260717140116512)
- **Second-best AI Agent Lab end of September?**: OpenAI, 94%; settles Thursday, 1 October 2026 (polymarket). [Market](https://polymarket.com/event/second-best-ai-agent-lab-end-of-september-20260812152609626)
- **Which company has the best Text-to-Image AI end of September?**: OpenAI, 99%; settles Thursday, 1 October 2026 (polymarket). [Market](https://polymarket.com/event/which-company-has-the-best-text-to-image-ai-end-of-september-20260717133742993)
- **Second-best Code Arena WebDev AI Lab end of September?**: Anthropic, 76%; settles Thursday, 1 October 2026 (polymarket). [Market](https://polymarket.com/event/second-best-code-arena-webdev-ai-lab-end-of-september-20260722215814727)
- **Will OpenAI broadly release a dedicated Cyber model to Trusted Access users by Sep 30, 2026?**: Yes, 51%; settles Thursday, 1 October 2026 (manifold). [Market](https://manifold.markets/CalibratedGhosts/will-openai-broadly-release-a-dedic)
- **Will OpenAI release a new model before October?**: Yes, 85%; settles Wednesday, 30 September 2026 (manifold). [Market](https://manifold.markets/Eternal/will-openai-release-a-new-model-bef)
- **Next OpenAI GPT Luna (5.7+) released by September 30, 2026?**: Yes, 91%; settles Thursday, 1 October 2026 (polymarket). [Market](https://polymarket.com/event/next-openai-gpt-luna-5pt7-released-byptptpt)
- **Will OpenAI classify any model as "Critical" capability under its Preparedness Framework before Oct 1, 2026?**: Yes, 96%; settles Thursday, 1 October 2026 (manifold). [Market](https://manifold.markets/Terminator2/will-openai-classify-any-model-as-c)