AI BriefingIntroducing celeris-1 (2 min read)
Key point
A next-generation language model, celeris-1, has been unveiled, applying Diffusion technology instead of the traditional Autoregressive approach to revolutionize speed.
Details
The new language model celeris-1 adopts a new inference architecture that uses Diffusion technology instead of the traditional Autoregressive generation method. This dramatically increases response speed while maintaining state-of-the-art levels of intelligence.
Key performance metrics are as follows:
- Response Speed: Achieves a p50 latency of 157ms, which is about 15x faster than GPT-5-mini and about 17x faster than GPT-5.
- Intelligence Level: Scores 76% on the MMLU-Pro benchmark, showing performance on par with GPT-5 models.
- Throughput: Processes 1,280 tokens per second, an overwhelming performance compared to gemini-3.5 flash-light (144 tokens/s).
celeris-1 is available for immediate use starting today, and you can apply for access through the official website.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Taste-Bench: Frontier LLMs Achieve 59.7% Accuracy on Decision Forking Tasks
Toss Builds 'Toss Benchmark' to Bridge Gap Between Public Leaderboards and Real-World Performance
JiT-DDT Architecture Improves Text-Image Model Training Speed by 3.6x Compared to Linum v2