New models

1. 16 October 2026

   Muse Wonder

   A typical tickets-test run cost an estimated $0.748. All 81 of its runs passed.

   Meta's Muse Wonder in Cursor CLI, four days after release, beside Claude Opus 5.5, GPT-6.1 Sol and Gemini 4 Argon, with costs at list prices.

2. 13 October 2026

   Gemini 4 Argon

   On 13 October's explain task, GPT-6.1 Sol and Gemini 4 Argon were close, and both usually finished ahead of Claude Opus 5.5.

   Gemini 4 Argon in Cursor CLI on its release day, beside Claude Opus 5.5 and GPT-6.1 Sol on the explain task.

3. 7 October 2026

   Claude Haiku 5.5 vs GPT-6 Luna

   At medium they were close, 7.3 s for Haiku 5.5 and 7.6 s for GPT-6 Luna. At xhigh GPT-6 Luna usually finished first, 9.0 s against Haiku 5.5's 15.2 s. One failed GPT-6 Luna run at medium is left out.

   Haiku 5.5 against GPT-6 Luna on the day Haiku came out: speed, real app work, skills and cost, at medium and xhigh effort.

4. 29 September to 6 October 2026

   GPT-6.1 Sol speed

   Every GPT-6.1 Sol run from 5 October's retest to 6 October streamed more text pieces a second than every run in the eight earlier sittings.

   GPT-6.1 Sol in Codex CLI, the same small job timed from launch day on, before and after OpenAI said it got faster.

The launch-day runs of Claude Sonnet 5.5 on 28 September and GPT-6.1 Sol on 29 September are in the Quick suite page's method.