Weekly Recap
The last week of 2025 brings its share of new features: Anthropic offers double usage for the holidays, Qwen launches a complete TTS suite with voice cloning, Mistral adds Skills to Vibe CLI, OpenAI details its security strategy for agents, and Gemini improves its photo editing capabilities.
Anthropic: 2x Usage for the Holidays
December 24-27, 2025 — Anthropic temporarily offers double usage limits to Pro and Max subscribers.
Starting at midnight PT tonight, all Pro and Max plans have 2x their usual usage limits through New Year’s.
🇬🇧 Starting at midnight PT tonight, all Pro and Max plans have 2x their usual usage limits through New Year’s. — @ClaudeAI on X
Details
| Plan | Normal Usage | Holiday Promo |
|---|---|---|
| Pro | Standard | 2x |
| Max | 5x | 10x |
Max 5x Fix
On December 27, Anthropic identified and fixed an issue affecting Max 5x users, resetting affected counters.
We’ve identified and fixed an issue with our usage promotion for Max 5x users. We’ve reset the affected counters.
🇬🇧 We’ve identified and fixed an issue with our usage promotion for Max 5x users. We’ve reset the affected counters. — @ClaudeAI on X
Qwen3-TTS: VoiceDesign and VoiceClone
December 23, 2025 — Alibaba unveils its new text-to-speech suite with two complementary models.
Meet the new Qwen3-TTS lineup: VoiceDesign & VoiceClone! Create, control, and clone voices—faster and more expressive than ever.
🇬🇧 Meet the new Qwen3-TTS lineup: VoiceDesign & VoiceClone! Create, control, and clone voices—faster and more expressive than ever. — @Alibaba_Qwen on X
VoiceDesign
Allows creating custom voices from text descriptions:
| Model | Latency | Use Case |
|---|---|---|
| VD-Flash | Ultra-fast | Streaming, real-time |
| VD-Turbo | Balanced | General production |
VoiceClone
Clones a voice from a few seconds of audio sample:
- High fidelity to original timbre
- Expressiveness preserved (tone, emotion)
- Multiple languages supported
Access
- Interactive Demo: Qwen Chat
- Official Blog: Qwen3-TTS VoiceDesign
- VoiceDesign: HuggingFace / ModelScope
- VoiceClone: HuggingFace / ModelScope
- API: VoiceDesign / VoiceClone
Qwen-Image-Edit-2511: Improved Image Editing
December 23, 2025 — Major update to Qwen’s image editing model.
Introducing Qwen-Image-Edit-2511 — a major upgrade over 2509, delivering significantly stronger consistency and more powerful real-world image editing.
🇬🇧 Introducing Qwen-Image-Edit-2511 — a major upgrade over 2509, delivering significantly stronger consistency and more powerful real-world image editing. — @Alibaba_Qwen on X
Improvements
| Aspect | 2509 → 2511 |
|---|---|
| Consistency | Better identity preservation |
| Real Editing | More precise on real photos |
| Quality | Reduced artifacts |
LightX2V Optimizations
The LightX2V framework brings Day-0 support with:
- 47% framework acceleration
- CFG + 4-step distillation: 25× compute reduction
- Ideal for production deployment
Availability
- Interactive Demo: Qwen Chat Image Edit
- Models: HuggingFace / ModelScope
- Demo: HuggingFace Spaces
Mistral: Skills for Vibe CLI
December 24, 2025 — Mistral extends its developer CLI with a system of reusable Skills.
We’ve shipped Skills for Vibe CLI: your dev expertise and rules, bundled and reusable across projects.
🇬🇧 We’ve shipped Skills for Vibe CLI: your dev expertise and rules, bundled and reusable across projects. — @MistralAI on X
Concept
Skills allow packaging:
- Project or team-specific coding rules
- Patterns and best practices
- Custom instructions for AI
Benefits
| Benefit | Description |
|---|---|
| Reusability | One skill works on all your projects |
| Consistency | Same standards applied everywhere |
| Sharing | Skills shareable between developers |
Context
This feature follows Vibe CLI improvements:
- December: Context 200k tokens, Devstral 2
- Now: Skills for customizing experience
OpenAI: ChatGPT Atlas Security
December 22, 2025 — OpenAI publishes an article detailing its strategy to protect against prompt injection for its agents.
We just published a post on how we continuously harden ChatGPT Atlas (and other agents) against prompt injection attacks.
🇬🇧 We just published a post on how we continuously harden ChatGPT Atlas (and other agents) against prompt injection attacks. — OpenAI Blog
Prompt Injection: The Agent Challenge
AI agents browsing the web or executing actions are vulnerable to malicious content hidden in visited pages.
OpenAI Strategy
| Layer | Protection |
|---|---|
| Detection | Identification of injection attempts |
| Isolation | Separation of system instructions and content |
| Monitoring | Continuous monitoring of abnormal behaviors |
Why It Matters
With the rise of autonomous agents (ChatGPT Atlas, Claude Computer Use, etc.), security becomes critical. This type of transparency helps the ecosystem progress collectively.
Gemini: Advanced Photo Editing
December 29, 2025 — Google demonstrates Gemini’s photo editing capabilities with a new feature.
Turn one image into an entire photoshoot. Edit the angle, lighting, lens and so much more.
🇬🇧 Turn one image into an entire photoshoot. Edit the angle, lighting, lens and so much more. — @GeminiApp on X
Features
From a single image, Gemini can generate variations with:
- Different angles (front, profile, overhead…)
- Varied lighting (studio, natural, dramatic…)
- Lens styles (wide angle, portrait, macro…)
- Multiple moods
Application
Particularly useful for:
- E-commerce: Creating multiple product views
- Creatives: Exploring variations quickly
- Social media: Generating varied content
Weekly Summary
| Date | Company | News |
|---|---|---|
| Dec 29 | Gemini multi-angle photo editing | |
| Dec 27 | Anthropic | Max 5x promo fix |
| Dec 24 | Anthropic | 2x usage Pro/Max holidays |
| Dec 24 | Mistral | Skills for Vibe CLI |
| Dec 23 | Qwen | Qwen3-TTS (VoiceDesign + VoiceClone) |
| Dec 23 | Qwen | Qwen-Image-Edit-2511 |
| Dec 22 | OpenAI | ChatGPT Atlas agent security |