SunBrief#87: Anthropic drops Claude Sonnet 5

Welcome to the SunBrief

Today in SunBrief 🌞

• Turn Your AI Idea Into a Working Product

• Anthropic Launches Claude Sonnet 5 as Its Most Agentic Model Yet

• Stock Updates

• Alibaba bans employees from using Claude Code

• Google Launches Nano Banana 2 Lite and Gemini Omni Flash

• AI Highlights of the Week

• Too Important to Miss

Turn Your AI Idea Into a Working Product

Digiden IT helps founders and companies turn AI concepts into working software, from product strategy and design to engineering, AI integration, and launch.

Whether you want to build an AI product, internal workflow tool, assistant, dashboard, or customer-facing AI product, Digiden IT builds around your exact use case without needing a full in-house engineering team.

We build:

• AI SaaS products
• AI copilots and assistants
• Internal AI tools
• AI dashboards and search systems
• AI automation tools

Go from AI idea to working product in as little as 45 days

Anthropic Launches Claude Sonnet 5 as Its Most Agentic Model Yet

New model brings stronger coding, tool use, reasoning, and autonomous task performance at a lower cost than Opus

Anthropic has introduced Claude Sonnet 5, its most agentic Sonnet model yet. It can plan, use tools like browsers and terminals, write code, and handle complex tasks more independently, bringing Sonnet closer to Opus-level performance at a lower price.

Key Points:

• Most agentic Sonnet yet: Sonnet 5 can plan, use tools, browse, run terminal tasks, and work more independently than earlier Sonnet models.

• Closer to Opus performance: Anthropic says Sonnet 5 narrows the gap with Opus 4.8, while offering better cost efficiency for many developers.

• Coding and tool-use upgrade: The model improves over Sonnet 4.6 in reasoning, software engineering, agentic search, computer use, and knowledge work.

• Available everywhere: Sonnet 5 is now the default model for Free and Pro users and is also available to Max, Team, Enterprise, Claude Code, and API users.

• Introductory pricing: Developers can use it at $2 per million input tokens and $10 per million output tokens through August 31, 2026, before standard pricing begins.

• Safer agentic behavior: Anthropic says Sonnet 5 has lower hallucination, sycophancy, and undesirable behavior rates than Sonnet 4.6.

• Lower cyber risk: Sonnet 5 performs far below Opus and Mythos models on dangerous cybersecurity tasks, though cyber safeguards are enabled by default.

Why It Matters:
Claude Sonnet 5 makes agentic AI more accessible for developers and teams, offering stronger coding, planning, tool use, and workflow support without Opus-level pricing. It also reflects Anthropic’s push to bring powerful AI agents into everyday work while keeping higher-risk cyber capabilities controlled.

Stock Updates

Alibaba bans employees from using Claude Code

Chinese tech giant shifts developers to its own Qoder assistant as Anthropic tightens access restrictions

Alibaba will reportedly ban employees from using Anthropic's Claude Code starting July 10, classifying the AI coding tool as high-risk software. The company is directing staff to use its in-house Qoder assistant instead, as Anthropic continues closing loopholes that allowed some Chinese users to access Claude.

Key Points:

• Internal ban: Alibaba will reportedly prohibit employees from using Claude Code beginning July 10.

• Shift to Qoder: Employees are being instructed to use Alibaba's own AI coding assistant, Qoder, instead of Anthropic's tool.

• High-risk classification: Alibaba has reportedly labeled Claude Code as high-risk software for internal use.

• Anthropic restrictions: The AI company already blocks Chinese companies and their overseas affiliates from accessing Claude models.

• Loophole crackdown: Anthropic has been strengthening measures to prevent unauthorized access, including an experimental system that could identify some Chinese users.

• Distillation concerns: Anthropic said the experiment was designed to stop unauthorized resellers and protect its models from AI distillation.

Why It Matters:

The reported ban reflects the growing split between Western and Chinese AI ecosystems. As access restrictions increase, Chinese companies are accelerating adoption of homegrown AI tools, while global AI providers strengthen controls over how and where their models can be used.

100+ ChatGPT Prompts to Revolutionize Your Day

Stop wasting time figuring out what to ask ChatGPT. This free HubSpot guide gives you 100+ proven prompts to speed up writing, research, planning, and daily tasks.

From better outputs to faster workflows, this is a simple shortcut to getting more value from ChatGPT. Instant access. Real use cases. Prompts you’ll actually use.

Google Launches Nano Banana 2 Lite and Gemini Omni Flash

New image and video models help developers build cheaper, faster, end-to-end creative workflows

Google DeepMind has launched Nano Banana 2 Lite and expanded developer access to Gemini Omni Flash, giving builders faster tools for image generation, video creation, and conversational editing across Google AI Studio, Gemini API, and Gemini Enterprise Agent Platform.

Key Points:

• Nano Banana 2 Lite: Google’s fastest and most cost-efficient Gemini image model, designed for high-volume image generation and rapid prototyping.

• Low-cost image generation: The model can generate images in around 4 seconds at about $0.034 per 1K image, making it useful for large-scale creative workflows.

• Upgrade path: Google recommends Nano Banana 2 Lite as the replacement for the original Nano Banana model, offering better quality, speed, and cost.

• Gemini Omni Flash: Google’s video generation and conversational editing model is now available to developers through the Gemini API and Google AI Studio.

• Video pricing: Omni Flash is priced at $0.10 per second of video output, matching Veo 3.1 Fast.

• Multimodal workflows: Developers can combine text, image, and video inputs, then use natural language to refine clips or create videos from generated images.

• Google ecosystem rollout: Nano Banana 2 Lite is also coming to Search AI Mode, Gemini app, NotebookLM, Google Photos, Stitch, Flow, and Google Ads.

• Safety layer: Both models use SynthID watermarking to help verify AI-generated content across Google products.

Why It Matters:
Google is making generative media faster, cheaper, and easier to build into real products. With image generation, video editing, and animation in one workflow, developers can move from concept images to finished clips without using separate tools.

AI Highlights of the Week

• OpenAI Introduces GeneBench-Pro for Biology AI

  OpenAI introduced GeneBench-Pro, a benchmark testing how AI agents handle messy biology data and judgment-heavy analysis.

  GPT-5.6 Sol scored up to 31.5% in Pro mode, showing progress but still clear limits in scientific reasoning.

• ElevenLabs Adds SynthID Watermarks to AI Audio

  ElevenLabs is adding SynthID watermarks to AI-generated audio, starting with free Text to Speech users before expanding further.

  The watermark stays through edits like trimming, speed changes, and file conversion, and can be checked with the free ElevenLabs Audio Detector.

• Acti Launches an Agentic Keyboard for AI Actions

  Acti introduced an agentic keyboard that turns text fields into an AI action layer, not just a place to type.

  Users can type an intent, hold the Acti Bar, and get actions like links, replies, summaries, or workflows without switching apps

• Anthropic Restores Claude Fable 5 Access

  Anthropic restored Claude Fable 5 after the U.S. lifted export controls on Fable 5 and Mythos 5.

  The new safety filter blocks the reported cyber issue in over 99% of cases and sends flagged requests to Opus 4.8 instead.

Too Important to Miss

Anthropic wants to develop its own drugs

Google Maps is building in-app food ordering

OpenAI 'In Early Talks to Give 5% Stake to US Governmentv

Meta Quietly Launches Pocket, an AI App for Creating Games

Last Week’s Poll Result

• Would you use AI as a medical copilot during a serious diagnosis?

  Yes, to organize and understand data → 86.96%

  Maybe, but only with doctor oversight → 13.04%

  

• What part of TikTok’s super app strategy is most powerful?

  TikTok Shop → 50.00%

  Payments and fintech → 50.00%

• Should AI companies block foreign labs accused of model extraction?

  Yes, immediately → 80.00%

  Maybe, after investigation → 20.00%

Feedback

Know someone who may be interested?