OpenAI releases low-resource language evaluation suite and GPT-6 series models
Current overview (not historical)
OpenAI released a new evaluation suite testing ChatGPT's performance in software engineering, SQLite, and low-resource programming languages such as COBOL. Subsequently, OpenAI launched the GPT-6 series models and reported results from the MultiLangBench test, claiming that GPT-6 Astra's average pass rate across 20 languages improved from approximately 38% for GPT-5.6 Terra to 87.4% under maximum reasoning effort. Reports indicate that GPT-6.1 Sol achieved 87% accuracy in legacy code migration under low reasoning effort and 98% under high reasoning effort, though experts warn significant barriers remain for actual code translation.
Written with AI from the reporting · updated 10/26/2026
Expand the detailed fact route and sources
Story route
A station-by-station route from the first public evidence onward; every station carries its own evidence and uncertainty.
- 1Unclassified fact stationOct 19, 00:00recorded Oct 20, 07:17single sourcerevised
OpenAI 发布低资源编程语言评估
Read original source ↗ OpenAI NewsReplay this moment in the explorerEvidence & revisions · 1 source records
- Subject：
- openai
- Action：
- 发布新评估
- Object：
- 软件工程、SQLite、COBOL迁移
- Occurred (fact side)：
- Oct 19, 00:00
- Published (valid side)：
- Oct 19, 20:00
- First seen (known side)：
- Oct 20, 07:17
Fact stationSingle source · one-sided claimRevisedCopy may have been revised later; time replay is bounded. Layer labels group reading only, not causation between layers.
- 2Unclassified fact stationOct 21, 00:00recorded Oct 21, 00:42single sourcerevised
OpenAI发布GPT-6多语言编程基准测试
Read original source ↗ The DecoderReplay this moment in the explorerEvidence & revisions · 1 source records
- Subject：
- openai
- Action：
- 发布GPT-6模型并公布MultiLangBench测试结果
- Object：
- GPT-6在COBOL、Fortran等低资源语言的编程能力
- Occurred (fact side)：
- Oct 21, 00:00
- Published (valid side)：
- Oct 21, 00:40
- First seen (known side)：
- Oct 21, 00:42
Fact stationSingle source · one-sided claimRevisedCopy may have been revised later; time replay is bounded. Layer labels group reading only, not causation between layers.
- 3Unclassified fact stationOct 26, 00:00recorded Oct 26, 21:17single sourcerevised
OpenAI报告称GPT-6.1 Sol在COBOL迁移任务中达87%行为保留率
Read original source ↗ The Verge · AIReplay this moment in the explorerEvidence & revisions · 1 source records
- Subject：
- openai
- Action：
- 发布博客报告
- Object：
- COBOL代码迁移到现代语言
- Occurred (fact side)：
- Oct 26, 00:00
- Published (valid side)：
- Oct 26, 21:15
- First seen (known side)：
- Oct 26, 21:17
Fact stationSingle source · one-sided claimRevisedCopy may have been revised later; time replay is bounded. Layer labels group reading only, not causation between layers.
Uncertainty marks flag where to check the sources; they are not proven causation.
Report timeline
Explore every side through dated source reports.
- The Verge · AIRelated fact: single sourceSource report revisedOpenAI claims AI can migrate legacy COBOL code to modern languages with improved accuracy
Report first recorded Oct 26, 21:17
OpenAI announced substantial improvements in its latest AI models' ability to migrate legacy software, such as COBOL, into modern languages like C++, Java, and Python. In internal tests, GPT-4.1 Sol achieved 87 percent behavior preservation accuracy at low reasoning effort and 98 percent at higher reasoning, potentially offering a solution for organizations maintaining critical infrastructure still reliant on decades-old code.
Read original source ↗
- The DecoderRelated fact: single sourceSource report revisedOpenAI claims GPT-6 significantly improves legacy programming language support
Report first recorded Oct 21, 00:42
OpenAI reports that its latest GPT-6 models substantially improve performance on legacy programming languages with limited training data, such as COBOL and Fortran. On the MultiLangBench evaluation, GPT-6 Astra achieved an average 87.4 percent test pass rate across 20 languages, up from about 38 percent for previous models. On a specific COBOL-to-code translation benchmark, GPT-6.1 Sol reached 86.6 percent accuracy at low reasoning effort and 97.9 percent with maximum reasoning effort, compared to single-digit success rates for GPT-5.6.
Read original source ↗
- OpenAI NewsFirst-party sourceRelated fact: single sourceSource report revisedOpenAI releases new evaluation suite for ChatGPT on low-resource programming languages
Report first recorded Oct 20, 07:17
OpenAI has introduced new evaluation benchmarks to assess ChatGPT's performance in low-resource programming languages, specifically targeting software engineering, SQLite operations, and COBOL migration tasks. These benchmarks aim to address the historical lack of standardized testing for languages with limited digital resources.
Read original source ↗
Event activity trend
- Latest comparable
- 3
- Comparable peak
- 3Oct 28, 14:00
- 24h change
- –
The chart compares consistently observed participants only. Hover or tap to inspect each hour; use arrow keys to move between observations.
Event discussion
Reader opinions are hosted on GitHub Discussions; posting requires a GitHub account. Comments are not verified facts.