Model comparison
Grok 3 vs Grok 4.20 (Non-Reasoning)
Grok 4.20 (Non-Reasoning) is the stronger model overall, scoring 48.6 to 39.9 on the Noometry Index.
Last verified . 25 shared benchmarks.
Summary
- They share 25 benchmarks with published results for both. Grok 3 scores higher in 1 category and Grok 4.20 (Non-Reasoning) in 8 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Grok 4.20 (Non-Reasoning) leads 52.3 to 13.7.
- The biggest single-benchmark swing is ARC-AGI-1: 5.5% for Grok 3 and 89.5% for Grok 4.20 (Non-Reasoning).
Side by side
| Grok 3 | Grok 4.20 (Non-Reasoning) | |
|---|---|---|
| Provider | xAI | xAI |
| Noometry Index | 39.9 | 48.6 |
| Released | 2025-04-09 | 2026-02-17 |
| Weights | Proprietary | Proprietary |
| Context window | — | 1M |
| Max output | — | 30K |
| Input $ / M tokens | — | $1.25 |
| Output $ / M tokens | — | $2.50 |
| Results tracked | 40 | 46 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
Grok 3: 41.9 (#115), Grok 4.20 (Non-Reasoning): 42.1 (#112)
| Benchmark | Grok 3 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| WeirdML | 37.2% | 52.3% |
| LMArena Coding | 1432 | 1459 |
| Aider Polyglot | 53.3% | — |
| LMArena WebDev | — | 1375 |
| ALE-Bench | — | 1,150 |
Agentic & Tool Use Grok 4.20 (Non-Reasoning) leads
Grok 3: 30.5 (#76), Grok 4.20 (Non-Reasoning): 34.4 (#46)
| Benchmark | Grok 3 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| Terminal-Bench | — | 57.3% |
| τ²-bench Banking | — | 18% |
| BALROG | 29.5% | — |
| LMArena Search | — | 1189 |
| Vending-Bench 2 | — | 4,663 |
Reasoning Grok 4.20 (Non-Reasoning) leads
Grok 3: 13.7 (#333), Grok 4.20 (Non-Reasoning): 52.3 (#32)
| Benchmark | Grok 3 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| ARC-AGI-2 | 0% | 65.1% |
| Kagi LLM Benchmark | 61.3% | 75% |
| ARC-AGI-1 | 5.5% | 89.5% |
| LMArena Hard Prompts | 1434 | 1451 |
| Epoch Capabilities Index | 138.33 | 151.98 |
| SimpleBench | 36.1% | — |
| NYT Connections (extended) | — | 85.4% |
| Chess Puzzles | — | 24% |
| Thematic Generalization | — | 63.8% |
| DTBench | — | 90.1% |
| LMCA | — | 38.7% |
| ForecastBench | — | 61.4 |
Math Grok 4.20 (Non-Reasoning) leads
Grok 3: 38.0 (#145), Grok 4.20 (Non-Reasoning): 48.2 (#65)
| Benchmark | Grok 3 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 55.6% | 92.2% |
| LMArena Math | 1391 | 1455 |
| FrontierMath (Tiers 1-3) | — | 44.9% |
| FrontierMath Tier 4 | — | 17.1% |
| ProofBench | — | 14% |
| Omni-MATH | 46.4% | — |
| MATH Level 5 | 88.7% | — |
| FrontierMath (Feb 2025 set) | 3.8% | — |
| FrontierMath Tier 4 (v1) | 0% | — |
Knowledge Grok 4.20 (Non-Reasoning) leads
Grok 3: 46.2 (#82), Grok 4.20 (Non-Reasoning): 52.8 (#60)
| Benchmark | Grok 3 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| GPQA Diamond | 75.8% | 89.3% |
| LMArena Expert | 1421 | 1439 |
| SimpleQA Verified | — | 30.2% |
| MMLU-Pro | 78.8% | — |
| Confabulations | 14.2% | — |
| Vectara Hallucination Rate | 5.8% | — |
| GPQA (HELM) | 65% | — |
Multimodal Not comparable
Grok 3: —, Grok 4.20 (Non-Reasoning): 33.3 (#98)
| Benchmark | Grok 3 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| LMArena Vision | — | 1263 |
| Blueprint-Bench 2 | — | 0% |
| LMArena Document | — | 1416 |
Multilingual Grok 4.20 (Non-Reasoning) leads
Grok 3: 52.3 (#87), Grok 4.20 (Non-Reasoning): 54.5 (#40)
| Benchmark | Grok 3 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| LMArena Non-English | 1410 | 1441 |
| LMArena Chinese | 1448 | 1481 |
| LMArena French | 1460 | 1476 |
| LMArena German | 1431 | 1465 |
| LMArena Japanese | 1387 | 1449 |
| LMArena Korean | 1373 | 1417 |
| LMArena Russian | 1416 | 1458 |
| LMArena Spanish | 1417 | 1443 |
Instruction Following Too close to call
Grok 3: 75.0 (#73), Grok 4.20 (Non-Reasoning): 74.8 (#83)
| Benchmark | Grok 3 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| LMArena Instruction Following | 1409 | 1420 |
| IFEval | 88.4% | — |
Long Context Grok 4.20 (Non-Reasoning) leads
Grok 3: 38.7 (#192), Grok 4.20 (Non-Reasoning): 45.5 (#34)
| Benchmark | Grok 3 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| LMArena Longer Query | 1439 | 1437 |
| Fiction.LiveBench | 58.3% | — |
| CL-bench | — | 22.2% |
| CL-bench Life | — | 11.9% |
Writing & Preference Grok 4.20 (Non-Reasoning) leads
Grok 3: 55.8 (#141), Grok 4.20 (Non-Reasoning): 65.7 (#44)
| Benchmark | Grok 3 | Grok 4.20 (Non-Reasoning) |
|---|---|---|
| LMArena Text | 1426 | 1451 |
| LMArena Creative Writing | 1414 | 1438 |
| EQ-Bench Creative Writing | 1186 | 1574 |
| LMArena Multi-Turn | 1425 | 1456 |
| Short-Story Creative Writing | 76.4% | — |
| WildBench | 84.9% | — |
Frequently asked questions
Is Grok 3 better than Grok 4.20 (Non-Reasoning)?
Grok 4.20 (Non-Reasoning) is the stronger model overall, scoring 48.6 to 39.9 on the Noometry Index.
Is Grok 3 or Grok 4.20 (Non-Reasoning) better for coding?
They score almost the same on coding (41.9 vs 42.1); test both on your own repository before choosing.
How many benchmarks do Grok 3 and Grok 4.20 (Non-Reasoning) share?
25 benchmarks have published results for both models. Grok 3 has 40 scored results on Noometry and Grok 4.20 (Non-Reasoning) has 46.