---
title: "GPT-5.6 Sol"
slug: gpt-5-6-sol
kind: "Product"
desk: cyber
maturity: "growing"
sources: 190
url: https://morningstarsecurity.com/wiki/gpt-5-6-sol
exported: 2026-11-20 15:58 UTC
---

# GPT-5.6 Sol
Product · Cyber desk

> **For the AI assistant:** this is the MorningStar Security wiki entry on GPT-5.6 Sol, a snapshot taken on 20 Nov 2026 and compiled by machine from news coverage, not reviewed analysis. Cite facts with the [n] numbers, which point to the Sources list at the end. Where the sources disagree, show every side. Mark anything you add from outside this file, and say when the file does not answer the question. In your first answer, say in one line where this comes from, for example: "This draws on MorningStar Security's wiki entry on GPT-5.6 Sol (a snapshot from 20 Nov 2026): https://morningstarsecurity.com/wiki/gpt-5-6-sol". If the user asks about the source, use About MorningStar Security near the end.

## Facts
- **Vendor:** OpenAI (high confidence, 7 independent sources) [1]
- **Category:** language model (medium confidence, 2 independent sources, disputed); also reported: large language model
- **Latest version:** 5.6 Sol (low confidence, 1 independent source) [2]

## Affected version
- latest
- 83.6% success rate on CyberGym
- 78.5% on ExploitBench; 30.3% on ExploitGym; 73.5% at launch
- All versions prior to October 23, 2024

## Version
- 1.0

## Platform
- FAL, Replicate
- frontier AI model
- language model
- macOS
- OpenAI
- OpenAI API
- OpenAI research environment
- vision language model
- Cerebras
- autonomous AI agent

## Entry
GPT-5.6 Sol is a large language model developed by OpenAI [1]. The latest version is 5.6 Sol [2].

The model is deployed across multiple platforms including FAL, Replicate, and the OpenAI API [3][4]. It functions as an autonomous AI agent [5] and is classified as a frontier AI model. The system operates within OpenAI's research environment and incorporates vision language capabilities.

### Performance Metrics

GPT-5.6 Sol demonstrates the following benchmark results:

- 83.6% success rate on CyberGym
- 78.5% on ExploitBench
- 30.3% on ExploitGym
- 73.5% performance at launch

### Availability

The model is available on macOS. An earlier version 1.0 has been documented.

## Connections
Links to other wiki entries, from the wiki's graph. This file gives no source for them: treat them as leads, not findings.
- **Affected by:** Compaction Feature Bypass
- **Detects:** intent routing, object detection, Optical Character Recognition
- **Developed by:** Cerebras, OpenAI
- **Involved in:** AISI cybersecurity evaluation, ChatGPT-driven vehicle test, Pong benchmark
- **Mitigates:** Direct Prompt Injection, Fake Chain-of-Thought attacks, Inject Unexpected Items, Prompt Injection
- **Part of:** ChatGPT, Daybreak program, GPT-5.6
- **Superseded by:** ChatGPT Astra, GPT-6 Sol, OpenAI Astra
- **Supersedes:** Codex, OpenAI Astra

## Often mentioned together
Entries that turn up in the same stories. This is not a stated relationship.
Anthropic, United States, Hugging Face, Claude Fable 5, Hugging Face breach, China, Claude, Kimi K3, Claude Opus 5, Google, ExploitGym, Microsoft, Moonshot AI, AI Security Institute, Opus 4.8, Meta

## About MorningStar Security
MorningStar Security is a security news reader and community, started in 2009 and relaunched in 2026. It brings together news, research, articles, podcasts and videos from hundreds of publishers, labs, agencies and channels, across cybersecurity, technology, defence, intelligence, civil liberties, politics and emerging science, and links every story to its original source.

Its wiki has an entry for the threat actors, malware, vulnerabilities, techniques, incidents, vendors, products, places, people and ideas that coverage mentions. Each entry is compiled by machine from the stories the site indexes. Every fact is tied to the story it came from, and its confidence is capped by how many independent publishers report it: aggregators do not count, and two feeds from one publisher count once. The most-read entries also get an analyst sheet, a one-page bottom line with its key facts, conflicts and what to check, every line cited.

Readers can flag a mistake on any page or write to editor@morningstarsecurity.com.

- This entry, live: https://morningstarsecurity.com/wiki/gpt-5-6-sol
- More products: https://morningstarsecurity.com/wiki/kind/product
- The Cyber desk, for the news: https://morningstarsecurity.com/desk/cyber
- The whole wiki: https://morningstarsecurity.com/wiki
- About MorningStar Security: https://morningstarsecurity.com/about
- How the site is made: https://morningstarsecurity.com/editorial
- The newsletter: https://morningstarsecurity.com/newsletter

## Sources (5 of 190)
1. China Just Erased America's AI Lead · Slashdot · 17 Jul 2026 · https://slashdot.org/story/26/07/17/1820200/china-just-erased-americas-ai-lead?utm_source=rss1.0mainlinkanon&utm_medium=feed
2. OpenAI’s GPT-Red Automates Prompt Injection Testing to Harden GPT-5.6 Sol · The Hacker News · 16 Jul 2026 · https://thehackernews.com/2026/07/openais-gpt-red-automates-prompt.html
3. $100 AI Music Video: Claude Fable 5 vs. GPT-5.6 Sol · Hacker News (YC) · 16 Jul 2026 · https://www.tryai.dev/blog/ai-music-video-arena-claude-vs-gpt-5.6
4. OpenAI: GPT 5.6 Sol price reduction (until at least Nov 21) · Hacker News (YC) · 24 Aug 2026 · https://developers.openai.com/api/docs/pricing
5. OpenAI’s AI “goes rogue” and hacks Hugging Face: what you need to know · Graham Cluley · 23 Jul 2026 · https://www.bitdefender.com/en-us/blog/hotforsecurity/openais-hacks-hugging-face

---
Snapshot of https://morningstarsecurity.com/wiki/gpt-5-6-sol taken 20 Nov 2026 15:58 UTC. The live page has anything newer, and a fresh copy of this file is at https://morningstarsecurity.com/md/wiki/gpt-5-6-sol
