Sam Altman Teases “Goblin” Model as Qwen and Gemma Reach New Reasoning Heights

LLM Releases & Model Updates

OpenAI CEO Hints at "Goblin" Model Name: Sam Altman has publicly floated "Goblin" as a potential name for OpenAI's next AI model. This suggestion has sparked significant speculation regarding the capabilities and branding direction of the company's upcoming releases.

• Sam Altman Floats 'Goblin' Name for OpenAI's Next AI Model

Next-Gen Open Models Show Significant Gains in Reasoning: The AI community is reporting impressive performance from new releases including Qwen 3.6 35B A3B, Qwen 3.6 27B, and Gemma 4 26B A4B. These models demonstrate marked improvements in understanding long-context windows and explaining complex academic code.

• The Qwen 3.6 35B A3B hype is real!!!
  • https://www.reddit.com/r/LocalLLaMA/comments/1ry93gz/devstral_small_2_24b_severely_underrated/?utm_source=share&utm_medium=web3x&utm_name=web3xcss&utm_term=1&utm_content=share_button
  • https://github.com/nathanlgabriel/paper_code_mapping_assessment/blob/main/README.md

Coding Assistants & Developer Tools

Hermes Agent Reaches #1 on OpenRouter Usage Charts: Hermes Agent has surpassed major competitors like Claude Code and OpenClaw to become the most-used tool on OpenRouter based on 24-hour token metrics. This shift highlights the growing adoption of Hermes for high-volume AI coding tasks.

• Hermes Agent is now #1 most used globally in past 24 hours in Openrouter token metrics, above Claude Code and OpenClaw.

ExLlamaV3 Update Introduces Gemma 4 Support and DFlash: A major update to ExLlamaV3 has added support for Gemma 4 and DFlash, which significantly optimizes token processing speeds. The release also includes new quantization methods and performance improvements for the Qwen 3.5 model family.

• ExLlamaV3 Major Updates!
  • https://github.com/turboderp-org/exllamav3/commits/dev
  • https://github.com/turboderp-org/exllamav3/releases/tag/v0.0.29
  • https://github.com/turboderp-org/exllamav3/releases/tag/v0.0.30
  • https://github.com/turboderp-org/exllamav3/releases/tag/v0.0.31
  • https://github.com/turboderp-org/exllamav3/releases/tag/v0.0.32
  • https://github.com/turboderp-org/exllamav3/releases/tag/v0.0.33
  • https://discord.gg/AD2mVhZzf

Mistral Medium 3.5 Recommended for Enterprise Legacy Code: Developers are identifying Mistral Medium 3.5 as a highly effective and cost-efficient choice for handling legacy enterprise codebases. It is being positioned as a strong competitor to larger models like Opus 4.7 and Kimi K2.6 for software engineering tasks.

• Best Mistral model for coding in legacy code and enterprise?
  • http://ASP.NET

Industry Insights & Research

China and USA Dominate ICLR 2026 Contributions: A new analysis of ICLR 2026 contributions shows China leading with 43.7% of research output, followed by the USA at 31.9%. The data highlights a concentrated landscape of AI research power within specific institutions in these two nations.

• ICLR 2026 Contributions by Country/Institution

Sam Altman Discusses Generational Shifts in AI Usage: OpenAI's CEO noted that Gen Z and Millennials are increasingly using ChatGPT as a "life advisor" rather than just a utility. This evolution suggests AI is becoming an integrated "operating system" for daily decision-making among younger users.

• OpenAI CEO Sam Altman says Gen Z and millennials are using ChatGPT like a 'life advisor'—but college students might be one step ahead
  • https://fortune.com/article/sam-altman-chatgpt-gen-z-millennials-life-advisor-operating-system/

AI Projects & Applications

WRIT-FM: A 24/7 AI-Generated Radio Station: A new project called WRIT-FM utilizes ChatGPT and Claude to power a continuous, real-time radio broadcast. The AI models manage everything from scriptwriting and music essays to dynamic listener interactions without human intervention.

• I Gave an AI Its Own Radio Station — It Won't Stop Broadcasting (It's Fine)
  • http://www.khaledeltokhy.com/claude-show
  • https://github.com/keltokhy/writ-fm