Model comparison
GPT-5.2 vs GPT-6 Astra
GPT-6 Astra is the stronger model overall, scoring 70.3 to 54.1 on the Noometry Index. GPT-5.2 costs 4.2× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Last verified . 40 shared benchmarks.
Summary
- They share 40 benchmarks with published results for both. GPT-5.2 scores higher in 0 categories and GPT-6 Astra in 10 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-6 Astra leads 85.1 to 50.1.
- The biggest single-benchmark swing is ProofBench: 15% for GPT-5.2 and 99% for GPT-6 Astra.
- GPT-5.2 is cheaper at $1.75 / $14 per million input/output tokens, against $10 / $50 for GPT-6 Astra.
- GPT-6 Astra accepts more context: 1.05M tokens versus 400K.
Side by side
| GPT-5.2 | GPT-6 Astra | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 54.1 | 70.3 |
| Released | 2025-12-11 | 2026-09-03 |
| Weights | Proprietary | Proprietary |
| Context window | 400K | 1.05M |
| Max output | 128K | 128K |
| Input $ / M tokens | $1.75 | $10 |
| Output $ / M tokens | $14 | $50 |
| Results tracked | 67 | 51 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-6 Astra leads
GPT-5.2: 52.2 (#32), GPT-6 Astra: 70.5 (#3)
| Benchmark | GPT-5.2 | GPT-6 Astra |
|---|---|---|
| LMArena WebDev | 1415 | 1792 |
| WeirdML | 72.2% | 93.6% |
| LMArena Coding | 1447 | 1488 |
| ALE-Bench | 1,294 | 2,951 |
| SWE-bench Verified | 73.8% | — |
| DeepSWE | — | 74.1% |
| FrontierCode | — | 53.3% |
| SWE-bench Verified (bash only) | 72.8% | — |
| SWE-bench Multilingual | 66.7% | — |
| FrontierSWE | — | 65.5% |
| SciCode | — | 56.5% |
| GSO | 27.4% | — |
| MirrorCode | — | 46.7% |
| AlgoTune | 2.05 | — |
Agentic & Tool Use GPT-6 Astra leads
GPT-5.2: 40.0 (#24), GPT-6 Astra: 52.3 (#3)
| Benchmark | GPT-5.2 | GPT-6 Astra |
|---|---|---|
| Remote Labor Index | 2.5% | 20.8% |
| Vending-Bench 2 | 3,591 | 15,515 |
| Terminal-Bench | 64.9% | — |
| APEX-Agents | — | 64.7% |
| Berkeley Function Calling Leaderboard | 55.9% | — |
| GDPval | 49.7% | — |
| τ²-bench Airline | 83% | — |
| τ²-bench Banking | 32.2% | — |
| τ²-bench Retail | 81.6% | — |
| τ²-bench Telecom | 89.7% | — |
| DeepResearch Bench | 41.1% | — |
| BALROG | — | 68.3% |
| LMArena Search | 1207 | — |
| METR Time Horizons | 75.3% | — |
Reasoning GPT-6 Astra leads
GPT-5.2: 50.1 (#33), GPT-6 Astra: 85.1 (#1)
| Benchmark | GPT-5.2 | GPT-6 Astra |
|---|---|---|
| ARC-AGI-2 | 52.9% | 95% |
| NYT Connections (extended) | 83.6% | 98.1% |
| ARC-AGI-1 | 86.2% | 98.5% |
| Chess Puzzles | 49% | 72% |
| EBR-Bench | 23% | 76.2% |
| LMArena Hard Prompts | 1445 | 1464 |
| Mystery Game Puzzles | 23% | 84% |
| DTBench | 90.9% | 97.3% |
| LMCA | 43.9% | 64.4% |
| Epoch Capabilities Index | 153.45 | 166.6 |
| SimpleBench | 45.8% | — |
| Kagi LLM Benchmark | 73.3% | — |
| CritPt | — | 31.7% |
| EnigmaEval | 10.4% | — |
| Bench to the Future 3 | — | 0.14 |
| ForecastBench | 60.1 | — |
Math GPT-6 Astra leads
GPT-5.2: 59.8 (#32), GPT-6 Astra: 93.4 (#1)
| Benchmark | GPT-5.2 | GPT-6 Astra |
|---|---|---|
| FrontierMath (Tiers 1-3) | 67.4% | 93.7% |
| FrontierMath Tier 4 | 31.7% | 97.6% |
| OTIS Mock AIME 2024-2025 | 96.1% | 100% |
| ProofBench | 15% | 99% |
| LMArena Math | 1439 | 1463 |
| MathArena Final-Answer Competitions | 72% | — |
| FrontierMath (Feb 2025 set) | 40.7% | — |
| FrontierMath Erdős | — | 2.9% |
| FrontierMath Tier 4 (v1) | 18.8% | — |
Knowledge GPT-6 Astra leads
GPT-5.2: 59.2 (#29), GPT-6 Astra: 75.3 (#1)
| Benchmark | GPT-5.2 | GPT-6 Astra |
|---|---|---|
| GPQA Diamond | 91.4% | 95.8% |
| Humanity's Last Exam | 27.8% | 54.8% |
| SimpleQA Verified | 37.1% | 75.6% |
| Vectara Hallucination Rate | 8.4% | 8.7% |
| LMArena Expert | 1445 | 1486 |
Multimodal GPT-6 Astra leads
GPT-5.2: 50.9 (#7), GPT-6 Astra: 55.1 (#3)
| Benchmark | GPT-5.2 | GPT-6 Astra |
|---|---|---|
| LMArena Vision | 1268 | 1280 |
| Furniture Assembly | 38.3% | 80% |
| LMArena Document | 1405 | 1468 |
| VPCT | 84% | — |
| Blueprint-Bench 2 | — | 49.7% |
Multilingual Too close to call
GPT-5.2: 53.6 (#63), GPT-6 Astra: 54.0 (#53)
| Benchmark | GPT-5.2 | GPT-6 Astra |
|---|---|---|
| LMArena Non-English | 1425 | 1431 |
| LMArena Chinese | 1460 | 1495 |
| LMArena French | 1456 | 1467 |
| LMArena Russian | 1440 | 1434 |
| LMArena Spanish | 1433 | 1412 |
| LMArena German | 1448 | — |
| LMArena Japanese | 1421 | — |
| LMArena Korean | 1390 | — |
Instruction Following GPT-6 Astra leads
GPT-5.2: 74.9 (#83), GPT-6 Astra: 76.6 (#37)
| Benchmark | GPT-5.2 | GPT-6 Astra |
|---|---|---|
| LMArena Instruction Following | 1417 | 1454 |
Long Context Too close to call
GPT-5.2: 44.0 (#74), GPT-6 Astra: 44.7 (#57)
| Benchmark | GPT-5.2 | GPT-6 Astra |
|---|---|---|
| LMArena Longer Query | 1428 | 1460 |
| CL-bench | 18.2% | — |
Writing & Preference GPT-6 Astra leads
GPT-5.2: 67.0 (#30), GPT-6 Astra: 75.7 (#7)
| Benchmark | GPT-5.2 | GPT-6 Astra |
|---|---|---|
| LMArena Text | 1439 | 1443 |
| LMArena Creative Writing | 1402 | 1423 |
| EQ-Bench Creative Writing | 1703 | 2173 |
| LMArena Multi-Turn | 1458 | 1457 |
Frequently asked questions
Is GPT-5.2 better than GPT-6 Astra?
GPT-6 Astra is the stronger model overall, scoring 70.3 to 54.1 on the Noometry Index. GPT-5.2 costs 4.2× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Which is cheaper, GPT-5.2 or GPT-6 Astra?
GPT-5.2 is cheaper. It lists at $1.75 per million input tokens and $14 per million output tokens; GPT-6 Astra lists at $10 and $50.
Is GPT-5.2 or GPT-6 Astra better for coding?
GPT-6 Astra scores higher on coding benchmarks: 70.5 versus 52.2 in the Noometry coding category.
Which has the bigger context window?
GPT-6 Astra does, with 1.05M tokens against 400K.
How many benchmarks do GPT-5.2 and GPT-6 Astra share?
40 benchmarks have published results for both models. GPT-5.2 has 67 scored results on Noometry and GPT-6 Astra has 51.