AI Daily · Oct 13
Oct 12 06:00 – Oct 13 06:00 (Beijing time) · 6 items
Updated 10/15 08:43 (Beijing time)
Today's highlights
- Related companies · OpenAIGPT-6.1 Sol Default Context Raised from 272k to 500kOpenAI announced that the default context for GPT-6.1 Sol in Codex and Work environments has been increased to 500k, up about 80% from the previous 272k, with a maximum support of 1M. Subscribers are exempt from long-context fees when deducting usage.
- Related companies · Alibaba QwenAlibaba Qwen releases QwenGyre, an online reinforcement learning framework for ultra-long-horizon agent tasksITHome reported that Alibaba Qwen released QwenGyre on Oct 12. It is an online reinforcement learning framework for long-horizon agent tasks. Through elastic scheduling and trajectory processing, the framework can turn ultra-long-horizon tasks into an effective training source and achieve speedup under the same GPU budget.
- Related companies · Zhipu AIZhipu GLM-5.3-Highspeed Rolls Out in Gray Release on ZcodeZhipu launched the new GLM-5.3-Highspeed model on Zcode. Subscribers are randomly given access to this gray-release version, and users have measured output speeds of up to 490 tokens/s.
Media reports
- Related companies
Alibaba QwenThird-party reportAlibaba Qwen releases QwenGyre, an online reinforcement learning framework for ultra-long-horizon agent tasks
ITHome reported that Alibaba Qwen released QwenGyre on Oct 12. It is an online reinforcement learning framework for long-horizon agent tasks. Through elastic scheduling and trajectory processing, the framework can turn ultra-long-horizon tasks into an effective training source and achieve speedup under the same GPU budget.
- Related companies
AnthropicThird-party reportClaude Opus 5.5 Max Tops Arena Agent Leaderboard
Arena released its Week 41, 2026 leaderboard, with Claude Opus 5.5 Max debuting at No. 1 on the agent leaderboard.
- Related companies
Zhipu AIThird-party reportZhipu GLM-5.3-Highspeed Rolls Out in Gray Release on Zcode
Zhipu launched the new GLM-5.3-Highspeed model on Zcode. Subscribers are randomly given access to this gray-release version, and users have measured output speeds of up to 490 tokens/s.
- Related companies
OpenAIThird-party reportGPT-6.1 Sol Default Context Raised from 272k to 500k
OpenAI announced that the default context for GPT-6.1 Sol in Codex and Work environments has been increased to 500k, up about 80% from the previous 272k, with a maximum support of 1M. Subscribers are exempt from long-context fees when deducting usage.
- Related companies
DeepSeekThird-party reportIntel CEO Lip-Bu Tan: Chinese models like DeepSeek are good enough and spreading fast
According to Reuters, Intel CEO Lip-Bu Tan said at a panel discussion organized by DBS Bank that Chinese AI models such as DeepSeek, Kimi, and GLM are already good enough. He noted that these models have lower token costs and are spreading very quickly.
- Related companies
Tencent HunyuanThird-party reportPony Ma Brings 28-Year-Old Talent to Discussion with Guangdong Party Secretary Huang Kunming: Tencent Chief AI Scientist Yao Shunyu Appears
According to China News Service, on Oct 8, Guangdong Party Secretary Huang Kunming visited Tencent headquarters. During the discussion, Tencent CEO Pony Ma introduced 28-year-old Chief AI Scientist Yao Shunyu. The report said Guangdong is betting on artificial intelligence, with a target of exceeding 300 billion yuan in the province's core industry scale by 2025.
Generates a long image you can save and share.