Google DeepMind·released 2026-09-301 source
Gemini 4 Argon benchmark scores: 12 benchmarks tracked, leading 4. 0% of its scores are independently reproduced — the rest are self-reported or unverified.
Leads on 4 of 12 benchmarks·best result 100 on OTIS Mock AIME 2024-2025 (unverified)·0 of 12 independently reproduced
Head-to-headGemini 4 Argon vs Claude Sonnet 5.5Leads on: APEX-Agents, GPQA diamond, OTIS Mock AIME 2024-2025, SimpleQA Verified
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| OTIS Mock AIME 2024-2025 | 100accuracy (%) | unverifiedT2 | 2026-09-30 |
| ProofBench | 99accuracy (%) | unverifiedT1 | 2026-09-30https://www.vals.ai/benchmarks/proof_bench |
| GPQA diamond | 95.2accuracy (%) | unverifiedT2 | 2026-09-30 |
| FrontierMath-Tiers-1-3-v2-Private | 85.96accuracy (%) | unverifiedT2 | 2026-09-30 |
| APEX-Agents | 82.4accuracy (%) | unverifiedT2 | 2026-09-30 |
| FrontierMath-Tier-4-v2-Private | 78.05accuracy (%) | unverifiedT2 | 2026-09-30 |
| SimpleQA Verified | 76accuracy (%) | unverifiedT2 | 2026-09-30 |
| Furniture Assembly | 60.71accuracy (%) | unverifiedT2 | 2026-09-30 |
| Mystery Game Puzzles | 60.34accuracy (%) | unverifiedT2 | 2026-09-30 |
| Chess Puzzles | 56.86accuracy (%) | unverifiedT2 | 2026-09-30 |
| FrontierSWE | 54.96accuracy (%) | unverifiedT1 | 2026-09-30https://www.frontierswe.com/ |
| FrontierCode | 47.55accuracy (%) | unverifiedT1 | 2026-09-30https://cognition.com/frontiercode |
Claims drawn from cited facts, not live model generation.
This model was developed by Google DeepMind and released in 2026. That recent vintage places this model in the newest generation of releases, arriving alongside the contemporary expectations and tooling of its cohort rather than those of an earlier era.
3 cited facts
This model is scored on 12 tracked benchmarks and currently holds top scores on 4 of them. Its record is unverified: outside evaluation has neither confirmed nor disputed the reported results.
3 cited facts
On average it trails the leading score by 8.47 points under the disclosed harness—a wide gap that reads as well behind the leaders rather than at the frontier. It nonetheless holds the top score on 4 benchmarks, so its standing is not a blanket absence of leadership but is concentrated in those categories.
2 cited facts
1 facts cross-checked across data sources: 1 single-source.
Source facts, citations, and refresh stamp for this record.
Sources: epoch_benchmarks
Gemini 4 Argon is an AI model developed by Google DeepMind, released 2026-09-30. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.
Gemini 4 Argon has recorded scores on 12 benchmarks, each shown with its evidence status.
Gemini 4 Argon has recorded scores on 12 benchmarks — GPQA diamond, OTIS Mock AIME 2024-2025, SimpleQA Verified, Chess Puzzles, FrontierMath-Tiers-1-3-v2-Private, FrontierMath-Tier-4-v2-Private, and 6 more. The full table above shows each score with its evidence status.
0 of 12 recorded scores (0%) are independently reproduced rather than self-reported by the lab.
Gemini 4 Argon has 12 tracked claims: 12 unverified.