Model comparison
GPT-5.6 Luna vs Grok 4.20 Multi-Agent
GPT-5.6 Luna is the stronger model overall, scoring 54.7 to 45.5 on the Noometry Index.
Last verified . 19 shared benchmarks.
Summary
- They share 19 benchmarks with published results for both. GPT-5.6 Luna scores higher in 8 categories and Grok 4.20 Multi-Agent in 1 category; 7 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.6 Luna leads 77.5 to 36.6.
- The biggest single-benchmark swing is NYT Connections (extended): 69.4% for GPT-5.6 Luna and 89.6% for Grok 4.20 Multi-Agent.
- GPT-5.6 Luna is cheaper at $0.20 / $1.20 per million input/output tokens, against $1.25 / $2.50 for Grok 4.20 Multi-Agent.
- GPT-5.6 Luna accepts more context: 1.05M tokens versus 1M.
Side by side
| GPT-5.6 Luna | Grok 4.20 Multi-Agent | |
|---|---|---|
| Provider | OpenAI | xAI |
| Noometry Index | 54.7 | 45.5 |
| Released | 2026-07-09 | 2026-03-09 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 1M |
| Max output | 128K | 30K |
| Input $ / M tokens | $0.20 | $1.25 |
| Output $ / M tokens | $1.20 | $2.50 |
| Results tracked | 51 | 20 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.6 Luna leads
GPT-5.6 Luna: 53.9 (#25), Grok 4.20 Multi-Agent: 42.4 (#91)
| Benchmark | GPT-5.6 Luna | Grok 4.20 Multi-Agent |
|---|---|---|
| LMArena Coding | 1466 | 1458 |
| DeepSWE | 67.2% | — |
| FrontierCode | 39.8% | — |
| CursorBench | 35.9% | — |
| LMArena WebDev | 1517 | — |
| SciCode | 53.6% | — |
| WeirdML | 60.9% | — |
| ALE-Bench | 1,667 | — |
Agentic & Tool Use Not comparable
GPT-5.6 Luna: 35.3 (#37), Grok 4.20 Multi-Agent: —
| Benchmark | GPT-5.6 Luna | Grok 4.20 Multi-Agent |
|---|---|---|
| BALROG | 45.6% | — |
| GDP.pdf | 22.7% | — |
| LMArena Search | — | 1204 |
| Vending-Bench 2 | 4,095 | — |
Reasoning GPT-5.6 Luna leads
GPT-5.6 Luna: 47.5 (#38), Grok 4.20 Multi-Agent: 42.4 (#44)
| Benchmark | GPT-5.6 Luna | Grok 4.20 Multi-Agent |
|---|---|---|
| NYT Connections (extended) | 69.4% | 89.6% |
| LMArena Hard Prompts | 1452 | 1448 |
| ARC-AGI-2 | 59.5% | — |
| SimpleBench | 46.8% | — |
| Kagi LLM Benchmark | 49.1% | — |
| ARC-AGI-1 | 88% | — |
| CritPt | 20.6% | — |
| Chess Puzzles | 40% | — |
| Mystery Game Puzzles | 21% | — |
| DTBench | 89.1% | — |
| LMCA | 48.5% | — |
| Surface Evolver Bench | 61.9% | — |
| Epoch Capabilities Index | 156.32 | — |
Math GPT-5.6 Luna leads
GPT-5.6 Luna: 77.5 (#12), Grok 4.20 Multi-Agent: 36.6 (#106)
| Benchmark | GPT-5.6 Luna | Grok 4.20 Multi-Agent |
|---|---|---|
| LMArena Math | 1455 | 1441 |
| FrontierMath (Tiers 1-3) | 82.1% | — |
| FrontierMath Tier 4 | 61% | — |
| OTIS Mock AIME 2024-2025 | 98.3% | — |
| ProofBench | 60% | — |
Knowledge GPT-5.6 Luna leads
GPT-5.6 Luna: 58.6 (#31), Grok 4.20 Multi-Agent: 39.7 (#116)
| Benchmark | GPT-5.6 Luna | Grok 4.20 Multi-Agent |
|---|---|---|
| LMArena Expert | 1479 | 1445 |
| GPQA Diamond | 91.6% | — |
| SimpleQA Verified | 41% | — |
Multimodal GPT-5.6 Luna leads
GPT-5.6 Luna: 43.0 (#22), Grok 4.20 Multi-Agent: 39.8 (#44)
| Benchmark | GPT-5.6 Luna | Grok 4.20 Multi-Agent |
|---|---|---|
| LMArena Vision | 1259 | 1259 |
| Blueprint-Bench 2 | 22.6% | — |
| Furniture Assembly | 42.5% | — |
| LMArena Document | 1457 | — |
Multilingual Grok 4.20 Multi-Agent leads
GPT-5.6 Luna: 53.0 (#73), Grok 4.20 Multi-Agent: 54.6 (#40)
| Benchmark | GPT-5.6 Luna | Grok 4.20 Multi-Agent |
|---|---|---|
| LMArena Non-English | 1418 | 1440 |
| LMArena Chinese | 1469 | 1475 |
| LMArena French | 1457 | 1467 |
| LMArena German | 1459 | 1456 |
| LMArena Japanese | 1416 | 1405 |
| LMArena Korean | 1414 | 1416 |
| LMArena Russian | 1428 | 1457 |
| LMArena Spanish | 1445 | 1447 |
Instruction Following Too close to call
GPT-5.6 Luna: 76.0 (#48), Grok 4.20 Multi-Agent: 75.0 (#80)
| Benchmark | GPT-5.6 Luna | Grok 4.20 Multi-Agent |
|---|---|---|
| LMArena Instruction Following | 1440 | 1420 |
Long Context Too close to call
GPT-5.6 Luna: 43.9 (#78), Grok 4.20 Multi-Agent: 43.7 (#84)
| Benchmark | GPT-5.6 Luna | Grok 4.20 Multi-Agent |
|---|---|---|
| LMArena Longer Query | 1437 | 1430 |
Writing & Preference GPT-5.6 Luna leads
GPT-5.6 Luna: 68.3 (#27), Grok 4.20 Multi-Agent: 64.1 (#57)
| Benchmark | GPT-5.6 Luna | Grok 4.20 Multi-Agent |
|---|---|---|
| LMArena Text | 1432 | 1450 |
| LMArena Creative Writing | 1397 | 1435 |
| LMArena Multi-Turn | 1433 | 1451 |
| EQ-Bench Creative Writing | 1829 | — |
| EQ-Bench 4 | 1156 | — |
Frequently asked questions
Is GPT-5.6 Luna better than Grok 4.20 Multi-Agent?
GPT-5.6 Luna is the stronger model overall, scoring 54.7 to 45.5 on the Noometry Index.
Which is cheaper, GPT-5.6 Luna or Grok 4.20 Multi-Agent?
GPT-5.6 Luna is cheaper. It lists at $0.20 per million input tokens and $1.20 per million output tokens; Grok 4.20 Multi-Agent lists at $1.25 and $2.50.
Is GPT-5.6 Luna or Grok 4.20 Multi-Agent better for coding?
GPT-5.6 Luna scores higher on coding benchmarks: 53.9 versus 42.4 in the Noometry coding category.
Which has the bigger context window?
GPT-5.6 Luna does, with 1.05M tokens against 1M.
How many benchmarks do GPT-5.6 Luna and Grok 4.20 Multi-Agent share?
19 benchmarks have published results for both models. GPT-5.6 Luna has 51 scored results on Noometry and Grok 4.20 Multi-Agent has 20.