Today, we are exiting stealth mode and launching Celtic Labs to the public.
For the past year, the AI industry has been locked in a race for scale. Companies are building bigger data centers, training on larger datasets, and burning billions of dollars to create models that are marginally better at standardized tests.
We believe they are optimizing for the wrong thing.
While models have become smarter at math, they have become worse at being human. They hedge. They moralize. They write in a flattened, corporate "average" voice. They have lost the spark.
Today, we are introducing Celt 4: The first frontier model family designed specifically for Creative Intelligence.
Introducing the Celt 4 Family
We are releasing two models today, each architected for a specific mode of cognition.
2. Celt 4 Lite: The Velocity
Intelligence at the Speed of Thought.
We believe high-quality intelligence should be abundant. Celt 4 Lite is our high-efficiency model, optimized for speed and daily utility. It outperforms GPT-4o-mini and Gemini Flash on constraint adherence and instruction following, while costing a fraction of the compute.
The Killer Feature: It is the default model for our Free and Pro tiers.
Specs: 1 Million Token Context. Sub-300ms latency.
Philosophy: Unlimited messages. No rationing intelligence.
The "Anti-Slop" Alignment
The most common complaint about modern AI is "AI Slop"—the verbose, empty language that screams "generated by a machine."
Celtic Labs has implemented a "Truth & Taste" alignment strategy.
- Truth: Our models perform an "Epistemic Audit" before answering, refusing to hallucinate libraries or facts.
- Taste: Our models are penalized during training for using generic filler words (delve, tapestry, symphony). The result is output that feels sharp, dense, and distinctly human.
Building for the Builders
We didn't build Celtic Labs for people who want to generate emails. We built it for the Creators—the engineers writing complex simulations, the writers crafting viral narratives, and the founders building the next unicorn.
We are offering a new standard of API access:
- Thought Access: Max users can see the raw <thinking> logs to understand how the model reached a conclusion.
- Strict JSON Mode: For reliable software integration.
- Fair Pricing: We are aggressively pricing our API to enable the next generation of startups.
Join the Frontier
This is Day 1.
We are not just launching a model; we are launching a movement back toward creativity, nuance, and ambition in AI.
We invite you to try Celt 4 Lite for free today, or upgrade to Max to experience the bleeding edge of creative reasoning.
Let's build something impossible.