Model comparison
GPT-5.4 vs Grok 4.20 (Non-Reasoning)
GPT-5.4 is the stronger model overall, scoring 59.4 to 48.6 on the Noometry Index. Grok 4.20 (Non-Reasoning) costs 3.6× less per token, which makes it the better buy when GPT-5.4's lead doesn't matter for your workload.
Last verified . 46 shared benchmarks.
Summary
- They share 46 benchmarks with published results for both. GPT-5.4 scores higher in 10 categories and Grok 4.20 (Non-Reasoning) in 0 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.4 leads 73.5 to 48.2.
- The biggest single-benchmark swing is ProofBench: 56% for GPT-5.4 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 $2.50 / $15 for GPT-5.4.
- GPT-5.4 accepts more context: 1.05M tokens versus 1M.
Side by side
| GPT-5.4 | Grok 4.20 (Non-Reasoning) | |
|---|---|---|
| Provider | OpenAI | xAI |
| Noometry Index | 59.4 | 48.6 |
| Released | 2026-03-05 | 2026-02-17 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 1M |
| Max output | 128K | 30K |
| Input $ / M tokens | $2.50 | $1.25 |
| Output $ / M tokens | $15 | $2.50 |
| Results tracked | 68 | 46 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.4 leads
GPT-5.4: 52.6 (#33), Grok 4.20 (Non-Reasoning): 42.1 (#112)
| Benchmark | GPT-5.4 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| LMArena WebDev | 1465 | 1375 |
| WeirdML | 77.7% | 52.3% |
| LMArena Coding | 1497 | 1459 |
| ALE-Bench | 1,607 | 1,150 |
| SWE-bench Verified | 76.9% | — |
| DeepSWE | 51.8% | — |
| SciCode | 56.6% | — |
| GSO | 31.4% | — |
| MirrorCode | 15.6% | — |
| AlgoTune | 1.85 | — |
Agentic & Tool Use GPT-5.4 leads
GPT-5.4: 46.5 (#13), Grok 4.20 (Non-Reasoning): 34.4 (#46)
| Benchmark | GPT-5.4 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| Terminal-Bench | 81.8% | 57.3% |
| τ²-bench Banking | 39.4% | 18% |
| LMArena Search | 1197 | 1189 |
| Vending-Bench 2 | 6,144 | 4,663 |
| APEX-Agents | 52.4% | — |
| DeepResearch Bench | 35.1% | — |
| PostTrainBench | 19% | — |
| GBAEval | 45.1% | — |
| METR Time Horizons | 74.3% | — |
Reasoning GPT-5.4 leads
GPT-5.4: 61.8 (#19), Grok 4.20 (Non-Reasoning): 52.3 (#32)
| Benchmark | GPT-5.4 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| ARC-AGI-2 | 74% | 65.1% |
| Kagi LLM Benchmark | 63.8% | 75% |
| NYT Connections (extended) | 91.3% | 85.4% |
| ARC-AGI-1 | 93.7% | 89.5% |
| Chess Puzzles | 44% | 24% |
| Thematic Generalization | 80% | 63.8% |
| LMArena Hard Prompts | 1485 | 1451 |
| DTBench | 94.4% | 90.1% |
| LMCA | 52% | 38.7% |
| Epoch Capabilities Index | 156.81 | 151.98 |
| ForecastBench | 59.5 | 61.4 |
| CritPt | 23.4% | — |
| EnigmaEval | 16% | — |
| EBR-Bench | 25.4% | — |
| Mystery Game Puzzles | 37% | — |
Math GPT-5.4 leads
GPT-5.4: 73.5 (#19), Grok 4.20 (Non-Reasoning): 48.2 (#65)
| Benchmark | GPT-5.4 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| FrontierMath (Tiers 1-3) | 78.6% | 44.9% |
| FrontierMath Tier 4 | 49% | 17.1% |
| OTIS Mock AIME 2024-2025 | 97.8% | 92.2% |
| ProofBench | 56% | 14% |
| LMArena Math | 1488 | 1455 |
| MathArena Final-Answer Competitions | 83.1% | — |
| FrontierMath (Feb 2025 set) | 47.6% | — |
| FrontierMath Tier 4 (v1) | 27.1% | — |
Knowledge GPT-5.4 leads
GPT-5.4: 65.3 (#14), Grok 4.20 (Non-Reasoning): 52.8 (#60)
| Benchmark | GPT-5.4 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| GPQA Diamond | 93.3% | 89.3% |
| SimpleQA Verified | 45.1% | 30.2% |
| LMArena Expert | 1507 | 1439 |
| Humanity's Last Exam | 36.2% | — |
| Vectara Hallucination Rate | 7% | — |
Multimodal GPT-5.4 leads
GPT-5.4: 43.7 (#20), Grok 4.20 (Non-Reasoning): 33.3 (#98)
| Benchmark | GPT-5.4 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| LMArena Vision | 1303 | 1263 |
| Blueprint-Bench 2 | 27.1% | 0% |
| LMArena Document | 1471 | 1416 |
| Furniture Assembly | 37.5% | — |
Multilingual GPT-5.4 leads
GPT-5.4: 56.2 (#23), Grok 4.20 (Non-Reasoning): 54.5 (#40)
| Benchmark | GPT-5.4 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| LMArena Non-English | 1465 | 1441 |
| LMArena Chinese | 1519 | 1481 |
| LMArena French | 1493 | 1476 |
| LMArena German | 1472 | 1465 |
| LMArena Japanese | 1485 | 1449 |
| LMArena Korean | 1448 | 1417 |
| LMArena Russian | 1480 | 1458 |
| LMArena Spanish | 1454 | 1443 |
Instruction Following GPT-5.4 leads
GPT-5.4: 77.1 (#27), Grok 4.20 (Non-Reasoning): 74.8 (#83)
| Benchmark | GPT-5.4 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| LMArena Instruction Following | 1469 | 1420 |
Long Context GPT-5.4 leads
GPT-5.4: 50.3 (#8), Grok 4.20 (Non-Reasoning): 45.5 (#34)
| Benchmark | GPT-5.4 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| CL-bench | 27.9% | 22.2% |
| CL-bench Life | 21.7% | 11.9% |
| LMArena Longer Query | 1473 | 1437 |
Writing & Preference GPT-5.4 leads
GPT-5.4: 71.9 (#17), Grok 4.20 (Non-Reasoning): 65.7 (#44)
| Benchmark | GPT-5.4 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| LMArena Text | 1469 | 1451 |
| LMArena Creative Writing | 1439 | 1438 |
| EQ-Bench Creative Writing | 1840 | 1574 |
| LMArena Multi-Turn | 1482 | 1456 |
| EQ-Bench 4 | 1272 | — |
Frequently asked questions
Is GPT-5.4 better than Grok 4.20 (Non-Reasoning)?
GPT-5.4 is the stronger model overall, scoring 59.4 to 48.6 on the Noometry Index. Grok 4.20 (Non-Reasoning) costs 3.6× less per token, which makes it the better buy when GPT-5.4's lead doesn't matter for your workload.
Which is cheaper, GPT-5.4 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-5.4 lists at $2.50 and $15.
Is GPT-5.4 or Grok 4.20 (Non-Reasoning) better for coding?
GPT-5.4 scores higher on coding benchmarks: 52.6 versus 42.1 in the Noometry coding category.
Which has the bigger context window?
GPT-5.4 does, with 1.05M tokens against 1M.
How many benchmarks do GPT-5.4 and Grok 4.20 (Non-Reasoning) share?
46 benchmarks have published results for both models. GPT-5.4 has 68 scored results on Noometry and Grok 4.20 (Non-Reasoning) has 46.