gpt-6IMAGE
GPT-6.1 Sol overtakes Astra on MathArena BrokenArXiv
On the math benchmark, 6.1 Sol scores 86.3% at less than half Astra's cost, and is nearly 9 points ahead of Opus 5.5 while being 2.6x cheaper.
Haider.
@haider1Archived 2026.10.01Originally posted 2026.09.30
Haider. @haider1Originally posted 2026.09.30
openai models remain undefeated in maths on MathArena BrokenArXiv: gpt-6.1 sol — 86.3% | $0.94 | 71.8k tokens gpt-6 astra — 81.9% | $2.26 | 34.0k opus 5.5 — 77.4% | $2.41 | 75.7k 6.1 sol leads astra while costing less than half as much, and is nearly 9 points ahead of opus 5.5 while being 2.6x cheaper
This post quotes another post: x.com/j_dekoninck/status/2105213644795523106
x.com/haider1/status/2105239984219488725Archived 2026.10.01 10:10