Megadose The AI curve, compressed.

GPT‑6.2 Astra and GPT‑6.2 Sol

Simon Willison ·
Astra’s input and output prices fall 20%, while cached input drops from $1 to $0.40 per million tokens.

Willison added support for both models in LLM 0.38; Sol retains its $2 input and $10 output pricing per million tokens. In his animation test, neither model convincingly transformed Yonkers into a giant mech. He describes the early reception as mixed, citing benchmark results below Fable 5.2 and complaints about confusing communication during long tasks.

Simon Willison's note

score 9

Categories: Model Releases