AI Daily: ChatGPT Pro, Claude Advisor Strategy, and Lyria 3 Pro for Free Users

ChatGPT Launches $100 Subscription, Claude Advisor Strategy Slashes Development Costs

Keeping a close eye on the tech world always reveals some truly impressive innovations. Today, several highly anticipated updates have arrived. From cost and efficiency concerns for developers to interactive entertainment for the general public, tech giants are showcasing their best work. Let’s dive in.

New OpenAI Pricing: $100/Month Pro Plan Targets Coding Needs

OpenAI recently officially updated the ChatGPT subscription mechanism. The original Plus plan seemingly couldn’t fully satisfy developers who heavily rely on AI for coding. Consequently, they introduced a brand-new Pro plan priced at $100 per month.

The highlight of this plan is providing five times the Codex usage compared to the Plus plan. This is fantastic news for users engaged in long-duration, high-intensity coding. Furthermore, the new plan retains all existing benefits, including unlimited access to Instant and Thinking models, and exclusively unlocks the GPT-5.4 Pro model for professional-grade reasoning tasks.

To celebrate the launch, OpenAI is offering a bonus: users who subscribe before May 31st can enjoy up to ten times the Codex limit. This isn’t just a numerical increase; it represents development teams being able to turn their wildest ideas into reality without hesitation.

Claude’s Clever Advisor Strategy: High Intelligence at an Affordable Price

How can AI agents be made smarter without breaking the bank? Anthropic’s latest blog post on “The advisor strategy” provides a perfect solution to this challenge.

The concept mirrors daily corporate operations. It assigns lower-cost models like Sonnet or Haiku the role of “executor,” responsible for handling mundane, sequential tasks such as tool calls and reading results. When the executor encounters a complex decision, it turns to the more powerful Opus model for help. Opus acts as the “advisor,” providing plans, corrections, or “stop” signals without performing the heavy lifting or generating content directly for the end-user.

Through this clever division of labor, the results are staggering: in the SWE-bench multilingual test, Sonnet paired with an Opus advisor increased scores by 2.7% while reducing per-task costs by 11.9%. Even more impressive, using the extremely low-cost Haiku as the executor with an Opus advisor saved up to 85% per task compared to running Sonnet alone, while more than doubling the performance score of Haiku alone.

Gemini’s Visual Evolution: Easily Generate Interactive Simulations

Technological progress isn’t limited to cold code. While users are accustomed to text or static images from chatbots, Google confirmed that the Gemini app now fully supports generating interactive simulations and 3D models.

What does this mean? It’s simple. If a user asks how the Moon orbits the Earth, the system no longer just provides a static diagram. The screen will display an interface where sliders can be manually adjusted, or different initial velocity or gravity values can be input, allowing users to see how those variables affect orbital stability. Google also emphasized its use in chemistry, such as allowing users to “rotate a molecule” to explore it deeply. This feature is now available to users worldwide; simply select the Pro model and use prompts like “help me visualize” to see complex concepts come to life.

Sparking Musical Inspiration: Lyria 3 Pro Now Available for Free Users with Longer, More Complex Tracks

Music creation has also received an exciting upgrade. Google announced today that the long-track music generation features of Lyria 3 Pro are now officially available to “Free users”! Users can now select “Create music” from the Gemini tools menu and choose the “Thinking” or “Pro” model to start creating for free. Unlike the standard Lyria 3, which generates 30 seconds, Lyria 3 Pro supports generating full tracks up to 3 minutes long.

This powerful tool introduces “Timestamp prompting,” allowing creators to precisely control transitions like a script, such as: “[00:00] Gospel choir opening, [00:15] add heavy bass hip-hop beat, [01:10] enter grand chorus”. It also supports multimodal generation, letting you upload up to 10 images or PDFs for the model to generate music based on the visual atmosphere. Additionally, Lyria 3 Pro supports clear vocal singing in 8 languages, including English, Japanese, and Korean, with built-in SynthID watermarking to ensure trust and security.

Waypoint-1.5 Arrives: Smooth Interactive Worlds on Standard GPUs

Finally, some “hardcore” but groundbreaking news. Generative virtual worlds have long required massive data center resources, but the Overworld team’s latest release, Waypoint-1.5, successfully breaks this hardware barrier.

This is a brand-new real-time interactive world model with 1.2 billion parameters (1.2B), trained on nearly 100 times more data than its predecessor. This massive increase in data allows the model to maintain physical consistency for 10 seconds (512 frames), a huge jump from the previous generation’s 2 seconds.

A technical article has also been published on Hugging Face. The team provides two versions: the Waypoint-1.5-1B 720p high-definition model for high-end devices, and a 360P lightweight version designed for broader accessibility. While optimized for consumer cards, the team still recommends NVIDIA cards with 16GB or more VRAM (RTX 5090 is recommended for a perfect 60 FPS experience). Users can download and run it via the Biome desktop app on GitHub. In the future, users will no longer need to rely on remote servers to explore smooth, real-time generated worlds on their own computers.

Example Videos:

Looking at today’s updates, we can see AI development entering a new phase. In the past, the focus was on parameter size and cloud computing power, but now giants are more concerned with “how to make powerful AI accessible, widespread, and cost-effective.” For businesses and developers, Claude’s advisor strategy is a prime example, perfectly demonstrating the wisdom of division of labor between large and small models (Opus as the brain, Haiku as the hands). For the general public, Google’s release of the professional Lyria 3 Pro music model to “Free users” instantly democratizes top-tier creative tools.

On the other hand, “interactivity” and “localization” are becoming new industry standards. Whether it’s Gemini allowing you to manually rotate molecules or Moon orbits, or Waypoint-1.5 breaking hardware barriers to run 3D generated worlds on consumer GPUs, AI is no longer just a “dialog box” spitting out text or images. It’s a “world” we can manipulate, explore, and direct. The next step for generative AI is clearly to let everyone become a creator of immersive experiences without any barriers.

Frequently Asked Questions (FAQ)

Q: How much is the new ChatGPT Pro plan, and what does it include? A: The new Pro plan is priced at $100 per month. In addition to 5x Codex limits and unlimited access to Instant/Thinking models, it exclusively unlocks the GPT-5.4 Pro model for professional reasoning.

Q: How exactly does Claude’s advisor strategy save costs? A: It uses large models (like Opus) for strategic guidance and smaller models (like Haiku or Sonnet) for actual execution. Since the expensive large model is only called for critical decisions, overall API costs are significantly reduced (e.g., Haiku with an Opus advisor can save up to 85% compared to running Sonnet alone).

Q: Does Gemini’s new interactive modeling feature require extra payment? A: This feature is currently open to all users worldwide. Simply select the Pro model in the prompt box and use keywords like “help me visualize” to start using interactive simulation tools, such as rotating molecular structures or changing physical variables.