PC Watch

Google released the image generation and editing AI model Nano Banana 2.1, claiming superior quality to top models at half the output cost.

• •Google released the image generation and editing AI model Nano Banana 2.1 on October 6.
• •It will be gradually provided in the Gemini app, Google AI Studio, Gemini API, Google Search AI Mode, Google Ads, Google Flow, and Google Stitch.
• •It is noted that cheaper models are beginning to outperform top-tier models.

Unitcom launched a limited-edition gaming PC with Ryzen and Radeon for 349,800 yen, commemorating an AMD award.

• •The model commemorates a double award, including the Radeon AI PRO Innovation Award, at AMD PARTNER ADVANCE 2026 JAPAN.
• •It features an all-AMD configuration with a Ryzen processor and Radeon graphics.
• •Priced at 349,800 yen, it is available in a limited run of 100 units starting October 6.

The ASUS Zenbook 14 with Snapdragon X X1-26-100 launched on October 7, showing 45% battery life after six hours of use and running Street Fighter 6 at nearly 60fps on low settings.

• •Street Fighter 6, which is not Arm-native, ran at nearly 60fps with low graphics settings.
• •The battery retained 45% charge after six hours of real-world use, including gaming.
• •The 14-inch Zenbook 14 UX3480QA with Snapdragon X X1-26-100 starts selling on October 7, priced from 189,800 yen.

llama.cpp v0.6.0 was released, adding support for the new GLM-5.3-Flash model and Metal-based acceleration.

• •ggml-org released llama.cpp v0.6.0 on October 5, which supports the new model 'GLM-5.3-Flash'.
• •The software is available for free download from GitHub and uses the MIT license.
• •Metal-based acceleration features were also added, further evolving local LLM execution environments.

OpenAI has sped up GPT-6 Astra and GPT-6.1 Sol by approximately 50%, reducing wait times for Codex users.

• •Developers using Codex via ChatGPT subscriptions can experience shorter response wait times without changing any settings.
• •On October 5, OpenAI announced it would implement daily improvements and usage limit resets for ChatGPT and Codex over the next 28 days.
• •This speedup was carried out as the first-day initiative within that 28-day improvement plan.

Report on trial results of the tool "Strata," which runs the 125B parameter "Qwen3.8-Flash-Next" model on a 12GB VRAM video card.

• •Information spread that "Qwen3.8-Flash-Next," with 125B total and 6B active parameters, runs on 12GB VRAM.
• •AI-related topics have been buzzing since October.
• •The model was actually run using "Strata" on a slightly older gaming PC environment.

Google replaced the "Gems" feature in the Gemini app with a new "Skills" feature, launching it globally on September 30.

• •Google announced on September 30 (US time) that it is replacing the "Gems" feature in the Gemini app with "Skills."
• •The "Skills" feature began rolling out globally in Gemini chats on the same day.
• •It is planned to be available for Google Workspace business, enterprise, non-profit, and education users within weeks.

Google announced that starting October 9, 2026, the models available in the Gemini app for individual users without AI subscriptions will be limited to Flash-Lite.

• •Google will change the model configuration for the Gemini app for individual accounts starting October 9, 2026.
• •The change applies to users who do not have an AI subscription.
• •Available models will be restricted to "Flash-Lite" only.

Magnitude AIがAIエージェント向けに開発したオープンソースのLLM推論エンジン「Magnitude」を公開した

• •AIエージェント向けに開発されたオープンソースのLLM推論エンジンである
• •ライセンスはApache-2.0であり、トークン費用がかからず完全無料で利用できる
• •ハードウェアごとに自動で最適化される機能を持つ

iPhoneをMacの外付けGPUのように活用してローカルAI処理を高速化するツール「Backburner」がGitHubで公開された

• •iPhoneをMacの外付けGPUのように活用し、ローカルAI処理を高速化するツール「Backburner」が公開された
• •同ツールはGitHubにて公開されている
• •利用するにはMacとiPhoneの両方にツールを導入する必要がある