OpenAI introduces ‘Ultrafast,’ a new mode that makes GPT-5.6 Sol work at 14x the speed

OpenAI is launching a preview of a sped up version of its latest, most powerful model, in an effort to court enterprise users.

The top 3

  1. Top 3 Fastest AI Models for Text Generation: While specific rankings vary by benchmark and task, Celeris-1 has reported impressive output speeds of 1,644 tokens/second, and Gemini 3.6 Flash achieves 238 tokens/second with a high quality score. For lowest latency to first answer, LiquidAI's LFM2-24B-A2B recorded 0.42 seconds.
  2. How Speed Boosts Real-World AI Applications: Faster AI models significantly enhance productivity and decision-making by automating repetitive tasks, analyzing vast data rapidly, and enabling real-time interactions in areas like customer service, medical diagnostics, and navigation.
  3. Key Metrics Defining AI Model Speed: AI model speed is primarily measured by Time to First Token (TTFT), which indicates initial response speed; Tokens Per Second (TPS), reflecting sustained generation rate; and End-to-End Latency (E2EL), which is the total time for a complete response.

Sources

Open the full topic