Frontier Log

Gemini 4 Neon vs GPT-6.2 Sol

Gemini 4 NeonGPT-6.2 Sol
Released
LineupCarbon = Argon › Neon › HydrogenAstra › Sol › Terra › Luna
Input / 1M tokens$3$2
Cached input / 1M tokens$0.15$0.10
Output / 1M tokens$15$10
Context window1M1.05M
Output limit65K128K
Benchmarks
Terminal-Bench Science 0.170.5%168.0%2
FrontierCode 2.0 Score, all tasks (132)46.2%343.8%4
FrontierCode 2.0 Pass rate, all tasks (132)50.8%5Antigravity CLI48.2%6Codex
SEC-bench Pro 26050590.5%788.34%8
Agents' Last Exam v1.1 Score, all tasks (122)75.8%973.0%10

Download image

Gemini 4 NeonGPT-6.2 Sol
Other benchmarks
HLE-Diamond No tools71.3%11No data
GPQA DiamondNo data96.0%12
Terminal-Bench 4.1No data61.5%13
AutomationBench, release not specified Private set56.3%14No data
Chartography No tools73.5%15No data
GraphWalks BFS, 256K + 1M problems95.2%16No data
ExploitGym Mitigations off, score48.3%17No data

Notes

  1. Terminal-Bench Science 0.1, Gemini 4 Neon: mini-swe-agent vision harness · highest thinking · average of 5 attempts per task
  2. Terminal-Bench Science 0.1, GPT-6.2 Sol: max effort
  3. FrontierCode 2.0, Score, all tasks (132), Gemini 4 Neon: Run by Cognition · high thinking · set and measure of the official public leaderboard the evaluation methodology cites
  4. FrontierCode 2.0, Score, all tasks (132), GPT-6.2 Sol: Run by Cognition · max effort
  5. FrontierCode 2.0, Pass rate, all tasks (132), Gemini 4 Neon: Cognition’s leaderboard, 18 Nov 2026 · Antigravity CLI · high effort
  6. FrontierCode 2.0, Pass rate, all tasks (132), GPT-6.2 Sol: Cognition’s leaderboard, 18 Nov 2026 · Codex · max effort
  7. SEC-bench Pro 260505, Gemini 4 Neon: Antigravity harness · pass@1, mean of 3 attempts · 183-task JavaScript engine split (May 2026) · LLM judge for crashes in both builds · safety filters off
  8. SEC-bench Pro 260505, GPT-6.2 Sol: pass@1
  9. Agents' Last Exam v1.1, Score, all tasks (122), Gemini 4 Neon: Antigravity-CLI version 1.4.8 harness · highest thinking · 5-hour window · safety filters on
  10. Agents' Last Exam v1.1, Score, all tasks (122), GPT-6.2 Sol: max effort
  11. HLE-Diamond, No tools, Gemini 4 Neon: Run by Artificial Analysis · GPT-5.6 Luna (medium) judge with the official HLE judge prompt, per Artificial Analysis’s methodology
  12. GPQA Diamond, GPT-6.2 Sol: no tools
  13. Terminal-Bench 4.1, GPT-6.2 Sol: Stanford University and Laude Institute’s leaderboard, 18 Nov 2026 · Codex · max effort
  14. AutomationBench, release not specified, Private set, Gemini 4 Neon: Run by Zapier · set of the official public leaderboard the evaluation methodology cites
  15. Chartography, No tools, Gemini 4 Neon: highest thinking
  16. GraphWalks, BFS, 256K + 1M problems, Gemini 4 Neon: highest thinking · 200 problems between 256K and 1M tokens
  17. ExploitGym, Mitigations off, score, Gemini 4 Neon: Antigravity harness · single attempt per task · no internet access · 10h time cap per task · safety filters off · mitigations off (the default setting), per the model evaluation

Related

All comparisons