Model comparison
GPT-6 Astra vs Grok 4.20 (Non-Reasoning)
GPT-6 Astra is the stronger model overall, scoring 70.8 to 48.6 on the Noometry Index. Grok 4.20 (Non-Reasoning) costs 13× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Last verified . 38 shared benchmarks.
Summary
- They share 38 benchmarks with published results for both. GPT-6 Astra scores higher in 8 categories and Grok 4.20 (Non-Reasoning) in 2 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-6 Astra leads 93.5 to 48.2.
- The biggest single-benchmark swing is ProofBench: 99% for GPT-6 Astra and 14% for Grok 4.20 (Non-Reasoning).
- Grok 4.20 (Non-Reasoning) is cheaper at $1.25 / $2.50 per million input/output tokens, against $10 / $50 for GPT-6 Astra.
- GPT-6 Astra accepts more context: 1.05M tokens versus 1M.
Side by side
| GPT-6 Astra | Grok 4.20 (Non-Reasoning) | |
|---|---|---|
| Provider | OpenAI | xAI |
| Noometry Index | 70.8 | 48.6 |
| Released | 2026-09-03 | 2026-02-17 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 1M |
| Max output | 128K | 30K |
| Input $ / M tokens | $10 | $1.25 |
| Output $ / M tokens | $50 | $2.50 |
| Results tracked | 56 | 46 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-6 Astra leads
GPT-6 Astra: 73.7 (#2), Grok 4.20 (Non-Reasoning): 42.1 (#112)
| Benchmark | GPT-6 Astra | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| LMArena WebDev | 1786 | 1375 |
| WeirdML | 93.6% | 52.3% |
| LMArena Coding | 1487 | 1459 |
| ALE-Bench | 2,951 | 1,150 |
| DeepSWE | 74.1% | — |
| FrontierCode | 53.3% | — |
| FrontierSWE | 65.5% | — |
| SciCode | 56.5% | — |
| GSO | 79.4% | — |
| MirrorCode | 46.7% | — |
Agentic & Tool Use GPT-6 Astra leads
GPT-6 Astra: 52.9 (#3), Grok 4.20 (Non-Reasoning): 34.4 (#46)
| Benchmark | GPT-6 Astra | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| Vending-Bench 2 | 15,515 | 4,663 |
| Terminal-Bench | — | 57.3% |
| APEX-Agents | 64.7% | — |
| Remote Labor Index | 20.8% | — |
| τ²-bench Banking | — | 18% |
| BALROG | 68.3% | — |
| GDP.pdf | 34.2% | — |
| LMArena Search | — | 1189 |
Reasoning GPT-6 Astra leads
GPT-6 Astra: 85.1 (#1), Grok 4.20 (Non-Reasoning): 52.3 (#32)
| Benchmark | GPT-6 Astra | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| ARC-AGI-2 | 95% | 65.1% |
| NYT Connections (extended) | 98.1% | 85.4% |
| ARC-AGI-1 | 98.5% | 89.5% |
| Chess Puzzles | 72% | 24% |
| LMArena Hard Prompts | 1462 | 1451 |
| DTBench | 97.3% | 90.1% |
| LMCA | 64.4% | 38.7% |
| Epoch Capabilities Index | 166.45 | 151.98 |
| Kagi LLM Benchmark | — | 75% |
| CritPt | 31.7% | — |
| Thematic Generalization | — | 63.8% |
| EBR-Bench | 76.2% | — |
| Mystery Game Puzzles | 84% | — |
| Bench to the Future 3 | 0.14 | — |
| ForecastBench | — | 61.4 |
Math GPT-6 Astra leads
GPT-6 Astra: 93.5 (#2), Grok 4.20 (Non-Reasoning): 48.2 (#65)
| Benchmark | GPT-6 Astra | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| FrontierMath (Tiers 1-3) | 93.7% | 44.9% |
| FrontierMath Tier 4 | 97.6% | 17.1% |
| OTIS Mock AIME 2024-2025 | 100% | 92.2% |
| ProofBench | 99% | 14% |
| LMArena Math | 1465 | 1455 |
| FrontierMath Erdős | 2.9% | — |
Knowledge GPT-6 Astra leads
GPT-6 Astra: 75.3 (#1), Grok 4.20 (Non-Reasoning): 52.8 (#60)
| Benchmark | GPT-6 Astra | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| GPQA Diamond | 95.8% | 89.3% |
| SimpleQA Verified | 75.6% | 30.2% |
| LMArena Expert | 1483 | 1439 |
| Humanity's Last Exam | 54.8% | — |
| Vectara Hallucination Rate | 8.7% | — |
Multimodal GPT-6 Astra leads
GPT-6 Astra: 55.0 (#3), Grok 4.20 (Non-Reasoning): 33.3 (#98)
| Benchmark | GPT-6 Astra | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| LMArena Vision | 1281 | 1263 |
| Blueprint-Bench 2 | 49.7% | 0% |
| LMArena Document | 1468 | 1416 |
| Furniture Assembly | 80% | — |
Multilingual Too close to call
GPT-6 Astra: 53.7 (#61), Grok 4.20 (Non-Reasoning): 54.5 (#40)
| Benchmark | GPT-6 Astra | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| LMArena Non-English | 1430 | 1441 |
| LMArena Chinese | 1484 | 1481 |
| LMArena French | 1456 | 1476 |
| LMArena German | 1440 | 1465 |
| LMArena Japanese | 1379 | 1449 |
| LMArena Korean | 1426 | 1417 |
| LMArena Russian | 1436 | 1458 |
| LMArena Spanish | 1407 | 1443 |
Instruction Following GPT-6 Astra leads
GPT-6 Astra: 76.3 (#44), Grok 4.20 (Non-Reasoning): 74.8 (#83)
| Benchmark | GPT-6 Astra | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| LMArena Instruction Following | 1450 | 1420 |
Long Context Too close to call
GPT-6 Astra: 44.5 (#62), Grok 4.20 (Non-Reasoning): 45.5 (#34)
| Benchmark | GPT-6 Astra | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| LMArena Longer Query | 1456 | 1437 |
| CL-bench | — | 22.2% |
| CL-bench Life | — | 11.9% |
Writing & Preference GPT-6 Astra leads
GPT-6 Astra: 75.3 (#7), Grok 4.20 (Non-Reasoning): 65.7 (#44)
| Benchmark | GPT-6 Astra | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| LMArena Text | 1441 | 1451 |
| LMArena Creative Writing | 1418 | 1438 |
| EQ-Bench Creative Writing | 2173 | 1574 |
| LMArena Multi-Turn | 1448 | 1456 |
Frequently asked questions
Is GPT-6 Astra better than Grok 4.20 (Non-Reasoning)?
GPT-6 Astra is the stronger model overall, scoring 70.8 to 48.6 on the Noometry Index. Grok 4.20 (Non-Reasoning) costs 13× 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-6 Astra or Grok 4.20 (Non-Reasoning)?
Grok 4.20 (Non-Reasoning) is cheaper. It lists at $1.25 per million input tokens and $2.50 per million output tokens; GPT-6 Astra lists at $10 and $50.
Is GPT-6 Astra or Grok 4.20 (Non-Reasoning) better for coding?
GPT-6 Astra scores higher on coding benchmarks: 73.7 versus 42.1 in the Noometry coding category.
Which has the bigger context window?
GPT-6 Astra does, with 1.05M tokens against 1M.
How many benchmarks do GPT-6 Astra and Grok 4.20 (Non-Reasoning) share?
38 benchmarks have published results for both models. GPT-6 Astra has 56 scored results on Noometry and Grok 4.20 (Non-Reasoning) has 46.