This Week in AI: The Cheap Tier Got Scary Good
GPT-6 reached every ChatGPT user, the small models got dramatically cheaper, and a wave of decision models arrived to do the boring calls.
The headline this week was reach, not raw power. OpenAI finished rolling GPT-6 out to Free and Go tiers around October 8, bringing the Intelligent UI, those interactive charts and buttons, to more than a billion weekly users. The model itself isn't new, but the surface is: ChatGPT now composes answers with tappable components instead of a wall of text.
On price, the small tiers did the real talking. Anthropic shipped Claude Haiku 5.5 on October 7 at $0.10 per million input tokens for prompts under 100,000 tokens, about 75 percent cheaper than the old Haiku, and halved Sonnet 5.5 cache-read pricing the same day. Google, meanwhile, limited free Gemini users to Flash Lite starting October 9, which is a quiet reminder that 'free' keeps shrinking at the top end.
The more interesting shift was the decision model. Perplexity's pplx-decider-v1-27b, out since October 1, costs $0.04 per million input tokens with free output. Cloudflare's Clef and Clef-flash, AWS's Strands Decider 2B, and Liquid AI's open d1 family all landed in early October to do one thing: return a calibrated choice instead of generating prose. OpenAI joined them on October 6 with a Decisions API that returns a fixed answer in roughly 150 milliseconds, and cut its API tiers from five to three the same day.
The thread is simple. Labs are racing on price and autonomy for the repeated, boring calls, not just on the flagship benchmark. For anyone picking a model this month, the right move is to test the cheap tier against whatever you run today before assuming you need the top shelf.
None of this is a reason to rewrite your stack. It's a reason to re-benchmark the calls you make a thousand times a day, because that's where the savings actually show up.