Gemini 3 Pro running SWE-agent on e2b via TGI 3.0, attempting AIME 2025.
resolvedReasoning
- 74.0
- score
- 3
- steps
- 1
- tool calls
- 3.4k
- tokens
- $0.06
- cost
- 1m
- wall time
Trace3 steps · 1 tool call · 3 rewarded
- 01thought
Let the roots be r, s, t. Vieta gives r+s+t = 12 and rs+st+tr = 47. We need r²+s²+t².
1160 tok2.2s - 02thought
r²+s²+t² = (r+s+t)² − 2(rs+st+tr) = 144 − 94 = 50.
1260 tok7.1sr=0.30 - 02tool call
pythonfrom sympy import symbols, solve x=symbols('x'); rs=solve(x**3-12*x**2+47*x-60, x); print(sum(r**2 for r in rs)) - 02observation
python50
r=0.20 - 03answer
050
950 tok0.9sr=1.00