AI Technology Observations | 10 October 2026, 07:00

This observation covers technical changes formally announced, made available or reliably verified in roughly the 24 hours to 7:00 am on 10 October 2026. Three items are included.

GPT-6.1 Sol Ultrafast becomes available

On 8 October, OpenAI moved Ultrafast mode for GPT-6.1 Sol from a future commitment to an active release across the API, Codex and ChatGPT Work. API pricing is US$12 per million input tokens and US$60 per million output tokens, while the official documentation provides WebSocket usage and an HTTP alternative. In Codex and ChatGPT Work, the mode is available on Pro 500 and eligible Enterprise and Edu plans. Enterprise access is off by default and must be enabled by a workspace administrator. Inference data residency is supported in the United States and Europe.

The descriptions of near-Astra intelligence and speeds of up to eight times Sol Standard are developer claims from OpenAI. There is no public independent reproduction of latency, throughput or task quality under uniform conditions. This change adds a higher-priced, lower-latency service tier to the same GPT-6.1 Sol model; it is not a new set of model weights or a separate model version.

ARTEX becomes closed source and ends public maintenance

CrowdStrike published its investigation of attacks on South Korean financial organisations on 7 October; the underlying activity took place from late September to early October. Researchers found Claude Code session histories, ARTEX configuration files and memory files in threat-actor-controlled open directories, leading them to assess that the actor used ARTEX and several external language models. CrowdStrike described the actor as likely Chinese-speaking and financially motivated, with moderate confidence. It did not attribute the activity to a named organisation, and the exact number of affected organisations remains unconfirmed.

The new fact in this observation is the ARTEX developer’s 8 October announcement that updates and maintenance would stop, no further versions would be released publicly, and the project would become closed source. Reuters subsequently verified that the GitHub page had been taken down. ARTEX had been an open-source automated penetration-testing agent that connected to external models including ChatGPT, Claude and DeepSeek; it was not a standalone foundation model. Removing the repository changes access to the code and future versions, but it cannot automatically recall historical copies that have already been downloaded or duplicated.

Report measures safety-evaluation disclosure for major Chinese AI model releases

SemiAnalysis published a research report on 8 October covering 857 identifiable model releases by nine major Chinese developers from 2021 to 15 September 2026, including 741 product models and 116 research models. Using a strict definition that required a result tied to a named model for harmful output, jailbreaks, privacy, refusal behaviour or dangerous capabilities, the researchers found that 31 releases, or 3.6 per cent, had ever been accompanied by a matching public safety result. Only nine, or 1.1 per cent, had a result available at or before release. No public safety disclosure was found for another 813 releases.

These figures measure public disclosure and do not establish that developers did not test models internally. The report also notes that different sizes and snapshots may be counted separately, so company-level rates are indicative rather than a ranking. It does not provide directly comparable figures for United States developers. The material is a research-firm report rather than a peer-reviewed paper; Reuters independently checked its sample scope, definitions and principal statistical findings.

Sources

OpenAI Developer Community, Ultrafast is rolling out today for GPT-6.1 Sol in the API, Codex, and ChatGPT Work, 8 October 2026, https://community.openai.com/t/ultrafast-is-rolling-out-today-for-gpt-6-1-sol-in-the-api-codex-and-chatgpt-work/1404475

OpenAI, ChatGPT & Codex changelog, 8 October 2026, https://learn.chatgpt.com/docs/changelog

CrowdStrike, Unknown Threat Actor Uses AI-Driven ARTEX to Target South Korean Finance, 7 October 2026, https://www.crowdstrike.com/en-us/blog/unknown-threat-actor-uses-artex-to-target-south-korean-finance/

Reuters, Chinese developer makes ARTEX AI agent closed-source after Korean bank hack, 9 October 2026, https://www.reuters.com/world/china/chinese-developer-makes-artex-ai-agent-closed-source-after-korean-bank-hack-2026-10-09/

SemiAnalysis, Beijing Will Not Pace the Frontier: China’s Speed-First AI Safety Regime, 8 October 2026, https://newsletter.semianalysis.com/p/beijing-will-not-pace-the-frontier

Reuters, China AI developers publish safety tests for just 3.6% of model releases, report finds, 9 October 2026, https://www.reuters.com/legal/litigation/china-ai-developers-publish-safety-tests-just-36-model-releases-report-finds-2026-10-09/