GPT-6.1 Sol vs Gemini 4 Argon
Google has officially announced Gemini 4 Argon, which beats GPT-6 models in certain benchmarks. Here is how the recently announced GPT-6.1 Sol compares against Gemini 4 Argon. Keep in mind that Gemini 4 Argon is not available to everyone yet, while GPT-6.1 Sol is already available.
Benchmarks
| Benchmark | GPT-6.1 Sol | Gemini 4 Argon | Difference |
|---|---|---|---|
| DeepSWE v1.1 | 75.2% | 77.9% | Argon +2.7 points |
| AutomationBench | 36.1% | 51.3% | Argon +15.2 points |
| Terminal-Bench Science 0.1 | 57.0% | 57.6% | Argon +0.6 points |
| OSWorld 2.0 (offline, partial credit) | 71.4% | 69.2% | GPT-6.1 Sol +2.2 points |
Artificial Analysis benchmarks
| Benchmark | GPT-6.1 Sol (Max) | Gemini 4 Argon (High) | Difference |
|---|---|---|---|
| Intelligence Index | 52 | 53 | Argon +1 |
| AA-Briefcase v1.1 | 1564 Elo | 1494 Elo | GPT-6.1 Sol +70 Elo |
| GDPval-AA v2.1 | 1575 Elo | 1611 Elo | Argon +36 Elo |
| AutomationBench-AA | 65% | 78% | Argon +13 points |
| Terminal-Bench 4.0 | 56% | 57% | Argon +1 point |
| SciCode | 54% | 62% | Argon +8 points |
| Humanity's Last Exam | 53% | 57% | Argon +4 points |
| GDP.pdf | 31% | 22% | GPT-6.1 Sol +9 points |
| CritPt | 32% | 27% | GPT-6.1 Sol +5 points |
| AA-Omniscience | 42 | 42 | Tie |
| AA-LCR v1.1 | 83% | 80% | GPT-6.1 Sol +3 points |
Cost and efficiency
| Metric | GPT-6.1 Sol (Max) | Gemini 4 Argon (High) | Difference |
|---|---|---|---|
| Intelligence Index | 52 | 53 | Argon +1 |
| Cost per Intelligence Index task | $0.72 | $1.99 | GPT-6.1 Sol about 64% cheaper |
| Output tokens across the Index | 67M | 110M | GPT-6.1 Sol uses about 39% fewer |
| Input, per 1M tokens | $2 | $2 | Same for now |
| Cached input, per 1M tokens | $0.10 | $0.10 | Same for now |
| Output, per 1M tokens | $10 | $10 | Same for now |
Pricing and specs
| Specification | GPT-6.1 Sol | Gemini 4 Argon | Difference |
|---|---|---|---|
| Input now, per 1M tokens | $2 | $2Introductory | Same |
| Cached input now, per 1M tokens | $0.10 | $0.10Introductory | Same |
| Output now, per 1M tokens | $10 | $10Introductory | Same |
| Input later, per 1M tokens | $2 | $4 | GPT-6.1 Sol 50% cheaper |
| Output later, per 1M tokens | $10 | $20 | GPT-6.1 Sol 50% cheaper |
| Context window | 1.05M tokens | About 1M tokens | Nearly identical |
| Max output | 128K tokens | 1M tokens | Argon about 7.8x larger |
| Text input | Yes | Yes | Same |
| Image input | Yes | Yes | Same |
| Public API availability | Available | Limited rollout at first | GPT-6.1 Sol is wider |