Amongst this week's model releases were Nvidia's Nemotron 3, OpenAI's ChatGPT Images and Google's Gemini Flash 3. Rumors are strong that GLM-4.7 and MiniMax 2.1 are imminent in their release.
wrt tooling, Wispr Flow is now offering in India Purchasing Power Parity pricing of ₹320 ($3.5) per month instead of $12 per month
Here's my referral link which gives you one month of free Pro subscription
Readers who think others in their family, friends and acquaintances who have a curiosity in knowing more about rapidly evolving AI tools/services/use cases would benefit from being subscribers of this weekly newsletter are encouraged to share this publication link to them and invite them to subscribe
The following messages were posted on the 'All Things AI' Telegram group from Sunday 14th Dec 2025 to Saturday 20th Dec 2025
Nvidia launches Nemotron 3, a family of AI models using a hybrid mixture-of-experts architecture and the Mamba-Transformer design, in 30B, 100B, and ~500B sizes
This 31.6 billion parameter model uses a mixture-of-experts design, activating just 3.6 billion per token for top efficiency. It handles a 1 million token context and excels on benchmarks like SWE-Bench Verified at 38.8% for coding and GPQA Diamond for reasoning, beating rivals while running faster. Developers can run it locally on 24GB RAM, with full datasets, recipes, and tools released openly for commercial use
Hybrid Mamba-Transformer architecture: Nemotron 3 Nano combines the hybrid Mamba-Transformer approach @NVIDIAAI has used on
Nemotron 3 Nano:
Up to 1M-token context
MoE: 31.6B total params / 3.6B active
Best-in-class SWE-Bench performance
Open weights + training recipe + redistributable datasets
And yes:
It is a small MoE reasoning model built for specialized agentic AI systems.
Just like others in the Nemotron family, Nano 3 is fully open with:
- Open weights, open data, & open recipes
- Designed for customization & optimization
https://www.wired.com/story/nvidia-becomes-major-model-maker-nemotron-3/
1 | From OpenAI Introducing ChatGPT Images, powered by our flagship new image generation model. - Stronger instruction following - Precise editing - Detail preservation - 4x faster than before Rolling out today in ChatGPT for all users, and in the API as GPT Image 1.5. |
2 | For folks based in India and looking for a good voice dictation which is reasonably priced Wispr Flow now has localized Indian pricing of ₹320 ($3.5) per month instead of $12 per month for annual billing. In effect using a form of PPP (Purchasing Power Parity) pricing model. Here's my referral link which gives you one month of free Pro subscription
Wispr Flow now has localized Indian Indian pricing of ₹320 ($3.5) per month instead of $12 per month for annual billing.
 |
3 | China's Xiaomi released MiMo-V2-Flash, an open-source Mixture-of-Experts LLM with 309B total parameters, 15B active parameters, and a 256k context window. The “hybrid thinking” switch controls whether it spends extra steps reasoning before answering, and the HTML generation and long context are aimed at agent loops where the model writes code, calls tools, reads outputs, and keeps going for hundreds of turns. Fuli Luo from the Xiaomi Mimo team is an ex-founding member of DeepSeek and core author of DeepSeek-V2.  Faster than Fast. Designed for Agentic AI.Introducing Xiaomi MiMo-V2-Flash — our new open-source MoE model: 309B total params, 15B active. Blazing speed meets frontier performance.  Highlights: Hybrid Attention: 5:1 interleaved 128-window SWA + Global | 256K context
MiMo-V2-Flash is live. It’s just step 2 on our AGI roadmap, but I wanted to dump some notes on the engineering choices that actually moved the needle.
Architecture: We settled on a Hybrid SWA. It’s simple, elegant, and in our internal benchmarks, it outperformed other Linear  China's Xiaomi released MiMo-V2-Flash, an open-source Mixture-of-Experts LLM with 309B total parameters, 15B active parameters, and a 256k context window.- Model on HuggingFace with MIT license - Beats DeepSeek-V3.2 on SWE-Bench Verified. - Hybrid Attention: 5:1 interleaved  |
Google Launches Gemini 3 Flash for Speed and Smarts
This new model tops benchmarks like GPQA Diamond at 90.4% and SWE-bench at 78%, beating Gemini 2.5 Pro and even Gemini 3 Pro in spots. It shines in coding, video analysis, and tool use, all at a quarter of the Pro model's cost—$0.50 per million input tokens. Leaders like Sundar Pichai and Demis Hassabis call it the best balance of power and efficiency, now live in apps, Search, and APIs for everyday and business tasks. Developers can build fast agentic workflows with it right away.
Gemini 3 Flash is our latest model with frontier intelligence built for lightning speed, and pushing the Pareto Frontier of performance and efficiency. It outperforms 2.5 Pro while being 3x faster at a fraction of the cost.
With this release, Gemini 3’s
Available in the API at $0.50 in / 1M tokens and $3.00 out / 1M tokens across.
Below is my personal website which aggregates links to many of my socials as well as content and community that I curate. Feel free to share this link to others who you think may find this content/community useful to them
The cover image of this newsletter via generated via the Seedream 4.5 model within the Freepik tool via the following prompt
lush forest, misty woods, ancient trees, verdant canopy, atmospheric light, woodland scenery, moss-covered stones, nature depth, cinematic forest lighting, enchanted woodland, soft fog, forestscape