Model comparison
GPT-4 vs GPT-6 Astra
GPT-6 Astra is the stronger model overall, scoring 70.3 to 28.9 on the Noometry Index.
Last verified . 23 shared benchmarks.
Summary
- They share 23 benchmarks with published results for both. GPT-4 scores higher in 0 categories and GPT-6 Astra in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-6 Astra leads 93.4 to 10.1.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 1.1% for GPT-4 and 100% for GPT-6 Astra.
- GPT-6 Astra is cheaper at $10 / $50 per million input/output tokens, against $30 / $60 for GPT-4.
- GPT-6 Astra accepts more context: 1.05M tokens versus 8K.
Side by side
| GPT-4 | GPT-6 Astra | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 28.9 | 70.3 |
| Released | 2023-03-14 | 2026-09-03 |
| Weights | Proprietary | Proprietary |
| Context window | 8K | 1.05M |
| Max output | 8K | 128K |
| Input $ / M tokens | $30 | $10 |
| Output $ / M tokens | $60 | $50 |
| Results tracked | 38 | 51 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-6 Astra leads
GPT-4: 31.2 (#277), GPT-6 Astra: 70.5 (#3)
| Benchmark | GPT-4 | GPT-6 Astra |
|---|---|---|
| WeirdML | 12.4% | 93.6% |
| LMArena Coding | 1254 | 1488 |
| DeepSWE | — | 74.1% |
| FrontierCode | — | 53.3% |
| LMArena WebDev | — | 1792 |
| FrontierSWE | — | 65.5% |
| SciCode | — | 56.5% |
| BigCodeBench Instruct | 46% | — |
| MirrorCode | — | 46.7% |
| BigCodeBench Complete | 57.2% | — |
| ALE-Bench | — | 2,951 |
| HumanEval+ | 79.3% | — |
Agentic & Tool Use Not comparable
GPT-4: —, GPT-6 Astra: 52.3 (#3)
| Benchmark | GPT-4 | GPT-6 Astra |
|---|---|---|
| APEX-Agents | — | 64.7% |
| Remote Labor Index | — | 20.8% |
| BALROG | — | 68.3% |
| METR Time Horizons | 36.1% | — |
| Vending-Bench 2 | — | 15,515 |
Reasoning GPT-6 Astra leads
GPT-4: 17.3 (#285), GPT-6 Astra: 85.1 (#1)
| Benchmark | GPT-4 | GPT-6 Astra |
|---|---|---|
| Chess Puzzles | 4% | 72% |
| LMArena Hard Prompts | 1241 | 1464 |
| Mystery Game Puzzles | 12% | 84% |
| DTBench | 62.7% | 97.3% |
| LMCA | 17.1% | 64.4% |
| Epoch Capabilities Index | 125.88 | 166.6 |
| ARC-AGI-2 | — | 95% |
| NYT Connections (extended) | — | 98.1% |
| ARC-AGI-1 | — | 98.5% |
| CritPt | — | 31.7% |
| EBR-Bench | — | 76.2% |
| Bench to the Future 3 | — | 0.14 |
| BIG-Bench Hard | 75.1% | — |
| ForecastBench | 57.8 | — |
| HellaSwag | 95.3% | — |
| WinoGrande | 87.5% | — |
Math GPT-6 Astra leads
GPT-4: 10.1 (#304), GPT-6 Astra: 93.4 (#1)
| Benchmark | GPT-4 | GPT-6 Astra |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.1% | 100% |
| LMArena Math | 1269 | 1463 |
| FrontierMath (Tiers 1-3) | — | 93.7% |
| FrontierMath Tier 4 | — | 97.6% |
| ProofBench | — | 99% |
| MATH Level 5 | 23% | — |
| FrontierMath Erdős | — | 2.9% |
| GSM8K | 92% | — |
Knowledge GPT-6 Astra leads
GPT-4: 18.4 (#277), GPT-6 Astra: 75.3 (#1)
| Benchmark | GPT-4 | GPT-6 Astra |
|---|---|---|
| GPQA Diamond | 35.7% | 95.8% |
| LMArena Expert | 1211 | 1486 |
| Humanity's Last Exam | — | 54.8% |
| SimpleQA Verified | — | 75.6% |
| Vectara Hallucination Rate | — | 8.7% |
| MMLU | 86.4% | — |
| TriviaQA | 84.8% | — |
Multimodal Not comparable
GPT-4: —, GPT-6 Astra: 55.1 (#3)
| Benchmark | GPT-4 | GPT-6 Astra |
|---|---|---|
| LMArena Vision | — | 1280 |
| Blueprint-Bench 2 | — | 49.7% |
| Furniture Assembly | — | 80% |
| LMArena Document | — | 1468 |
Multilingual GPT-6 Astra leads
GPT-4: 40.8 (#210), GPT-6 Astra: 54.0 (#53)
| Benchmark | GPT-4 | GPT-6 Astra |
|---|---|---|
| LMArena Non-English | 1246 | 1431 |
| LMArena Chinese | 1242 | 1495 |
| LMArena French | 1282 | 1467 |
| LMArena Russian | 1251 | 1434 |
| LMArena Spanish | 1261 | 1412 |
| LMArena German | 1251 | — |
| LMArena Japanese | 1209 | — |
| LMArena Korean | 1184 | — |
Instruction Following GPT-6 Astra leads
GPT-4: 65.5 (#217), GPT-6 Astra: 76.6 (#37)
| Benchmark | GPT-4 | GPT-6 Astra |
|---|---|---|
| LMArena Instruction Following | 1241 | 1454 |
Long Context GPT-6 Astra leads
GPT-4: 37.7 (#207), GPT-6 Astra: 44.7 (#57)
| Benchmark | GPT-4 | GPT-6 Astra |
|---|---|---|
| LMArena Longer Query | 1244 | 1460 |
Writing & Preference GPT-6 Astra leads
GPT-4: 35.0 (#263), GPT-6 Astra: 75.7 (#7)
| Benchmark | GPT-4 | GPT-6 Astra |
|---|---|---|
| LMArena Text | 1263 | 1443 |
| LMArena Creative Writing | 1244 | 1423 |
| EQ-Bench Creative Writing | 752 | 2173 |
| LMArena Multi-Turn | 1257 | 1457 |
Frequently asked questions
Is GPT-4 better than GPT-6 Astra?
GPT-6 Astra is the stronger model overall, scoring 70.3 to 28.9 on the Noometry Index.
Which is cheaper, GPT-4 or GPT-6 Astra?
GPT-6 Astra is cheaper. It lists at $10 per million input tokens and $50 per million output tokens; GPT-4 lists at $30 and $60.
Is GPT-4 or GPT-6 Astra better for coding?
GPT-6 Astra scores higher on coding benchmarks: 70.5 versus 31.2 in the Noometry coding category.
Which has the bigger context window?
GPT-6 Astra does, with 1.05M tokens against 8K.
How many benchmarks do GPT-4 and GPT-6 Astra share?
23 benchmarks have published results for both models. GPT-4 has 38 scored results on Noometry and GPT-6 Astra has 51.