Up to four models
GPT-6 Astra vs Gemini 4 Argon
On the index (GPT-4 = 100), GPT-6 Astra scores 199 and Gemini 4 Argon 195. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
GPT-6 Astra
199182–215
WeightsClosed
ReleasedSep 2026
Price in / out$10 / $50
Context1.05M
Core tests taken7 of 9
Gemini 4 Argon195179–212
WeightsClosed
ReleasedSep 2026
Price in / out$2.0 / $10
Context1.05M
Core tests taken6 of 9
GPT-6 Astra leads Gemini 4 Argon by 3 points, ahead on 3 of 6 shared core tests, at 5.0× the price.
Test by testthird-party results
GPT-6 AstraGemini 4 Argon
Gemini 4 ArgonEvery result as a table
| Test | GPT-6 Astra | Gemini 4 Argon |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 95.8% | 96.4% |
| EBR-bench | 76.2% | 64.8% |
| SimpleQA Verified | 75.6% | 76.0% |
| Maths | ||
| FrontierMath (tiers 1-3) | 93.7% | 86.0% |
| AIME-style maths (OTIS mock) | 100.0% | 100.0% |
| ProofBench v1.1 | 99.0% | 99.0% |
| Chess puzzles | 74.0% | 59.0% |
| FrontierMath (tier 4) | 97.6% | 78.0% |
| Coding | ||
| Terminal-Bench 4.0 | 59.6% | 57.6% |
| Code Migration | 67.7% | 68.2% |
| CyberBench v1.1 | 41.1% | 77.9% |
| IOI | 100.0% | 100.0% |
| ProgramBench | 5.5% | 2.5% |
| Vibe Code Bench 1-100 | 27.6% | 21.2% |
| Vibe Code Bench v1.1 | 89.6% | 91.9% |
| FrontierSWE | 65.5% | 55.0% |
| Agents | ||
| APEX-Agents | 64.7% | 82.4% |
| BioMysteryBench | 79.3% | 76.3% |
| SRE Bench | 56.9% | 44.3% |
| Terminal-Bench-Science | 62.9% | 44.3% |
| DTBench | 97.3% | 97.6% |
| Furniture assembly | 80.0% | 72.5% |
| LMCA | 64.4% | 63.7% |
| Mystery game puzzles | 84.0% | 64.0% |
| Professional work | ||
| Finance Agent (v2) | 53.5% | 65.4% |
| EMB | 71.7% | 75.2% |
| HLAB-Vals-1.1 | 7.1% | 27.1% |
| Legal Research Bench | 39.4% | 54.8% |
| MedCode | 48.5% | 58.8% |
| MedScribe | 87.9% | 87.4% |
| SAGE | 46.4% | 53.6% |
| Tax Agent Bench | 20.7% | 44.6% |
| Novel problems | ||
| ARC-AGI-2 | 95.0% | 91.3% |
| ARC-AGI-1 | 98.5% | 98.5% |
| ARC-AGI-3 | 62.7% | 49.8% |
| MysteryMechanism | 53.1% | 45.5% |
Tests at least two of these models have taken. Bold rows are the core tests.
