Hello from The AI Night,
Today in AI:
OpenAI Set to Begin Rolling Out Ads in ChatGPT
OpenAI Rolling Out "ChatGPT Go" Globally at $8/month
Google DeepMind launches TranslateGemma: Open Translation Models in 4B,12B and 27B Parameters
Source: OpenAI blog
Here's the deal: OpenAI announced it will begin testing advertisements in ChatGPT in the coming weeks for U.S. users on the free and new Go tiers. The company also launched ChatGPT Go globally at $8/month with expanded features including image creation and memory.
The Breakdown:
Ads will appear at the bottom of responses when a relevant sponsored product exists based on the current conversation
All ads will be clearly labeled and separated from ChatGPT's actual answers
Pro, Business and Enterprise subscriptions remain ad free
Users under 18 will not see ads; sensitive topics like health and politics are excluded
OpenAI states ads will never influence ChatGPT's responses and user data will not be sold to advertisers
Users can turn off personalization and clear ad related data at any time
ChatGPT Go has been available in 171 countries since August and now expands to the U.S.
The bigger picture: This marks OpenAI's first move into advertising revenue. The approach creates a clear separation between paid “ad free tiers” and “ad supported” free access. It signals a diversified business model beyond subscriptions while testing whether ads can coexist with user trust in AI assistants.
Source: OpenAI blog
Here's the deal: OpenAI has rolled out ChatGPT Go worldwide after an initial launch in India in August 2025. The plan costs $8/month in the US with localized pricing in other markets.
The Breakdown:
Go provides 10x more messages, file uploads and image creation than the free tier
Users get access to GPT-5.2 Instant with longer memory and context windows
Three consumer tiers now exist: Go ($8), Plus ($20) and Pro ($200)
Plus offers GPT-5.2 Thinking and the Codex coding agent for deeper reasoning tasks
Pro unlocks GPT-5.2 Pro with maximum memory and early feature access
OpenAI will begin testing ads in the free tier and Go plan in the US soon
Plus, Pro, Business and Enterprise plans remain ad free
The bigger picture: This creates a clear middle ground between free users and power users. The ad-supported model signals OpenAI is prioritizing scale and accessibility while reserving premium experiences for paying subscribers.
Google DeepMind
Google DeepMind launches TranslateGemma: Open Translation Models in 4B,12B and 27B Parameters
Source: Google blog
Here's the deal: Google DeepMind released TranslateGemma, a suite of open translation models in 4B, 12B, and 27B parameter sizes, built on Gemma 3 and distilled from Gemini. The models cover 55 evaluated languages with training on nearly 500 additional language pairs.
The Breakdown:
The 12B model outperforms the Gemma 3 27B baseline on WMT24++ benchmark, same quality at less than half the parameters
Two stage training: supervised fine-tuning on human + synthetic parallel data, then reinforcement learning using MetricX-QE and AutoMQM reward models
4B variant targets mobile/edge deployment; 12B runs on consumer laptops; 27B fits on a single H100
Multimodal capabilities retained can translate text in images without additional fine tuning
Available now on Hugging Face, Kaggle and Vertex AI
The bigger picture: This signals continued institutional confidence in frontier AI labs despite bubble concerns. For Anthropic specifically it provides a runway for compute and talent while keeping IPO optionality open.
What else you need to know:
Higgsfield raised $80 million at a $1.3 billion valuation for its AI video generation platform, reporting $200 million in annualized revenue with social media marketers driving 85% of usage.
DEV and Google AI launched a portfolio challenge through February 1 offering three $1,000 prizes for developers who build portfolios using Gemini tools deployed on Cloud Run.
Wikimedia Enterprise added Amazon, Meta, Microsoft, Mistral AI and Perplexity as partners alongside Google and others paying for reliable API access to Wikipedia content for AI systems.
Black Forest Labs released FLUX.2 [klein], a compact image model family delivering sub-second generation and editing on consumer GPUs with 13GB VRAM with the 4B variant under Apache 2.0.
German AI customer service startup Parloa raised $350 million in Series D funding led by General Catalyst, tripling its valuation to $3 billion from $1 billion in July 2025.
That’s it for today’s edition of The AI Night.
Our goal is to cut through the noise, surface what actually changed, and explain why it matters.
If this was useful, you’ll get the same signal here tomorrow.