Model comparison
Grok 4.20 Multi-Agent vs Kimi K2.5
Kimi K2.5 is the stronger model overall, scoring 48.1 to 46.2 on the Noometry Index.
Last verified . 19 shared benchmarks.
Summary
- They share 19 benchmarks with published results for both. Grok 4.20 Multi-Agent scores higher in 2 categories and Kimi K2.5 in 7 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Kimi K2.5 leads 53.6 to 40.4.
- The biggest single-benchmark swing is NYT Connections (extended): 89.6% for Grok 4.20 Multi-Agent and 69.9% for Kimi K2.5.
- Kimi K2.5 is cheaper at $0.45 / $2.25 per million input/output tokens, against $1.25 / $2.50 for Grok 4.20 Multi-Agent.
- Grok 4.20 Multi-Agent accepts more context: 1M tokens versus 262K.
- Kimi K2.5 has downloadable open weights; the other is API-only.
Side by side
| Grok 4.20 Multi-Agent | Kimi K2.5 | |
|---|---|---|
| Provider | xAI | Moonshot AI |
| Noometry Index | 46.2 | 48.1 |
| Released | 2026-03-09 | 2026-01-27 |
| Weights | Proprietary | Open |
| Context window | 1M | 262K |
| Max output | 30K | 262K |
| Input $ / M tokens | $1.25 | $0.45 |
| Output $ / M tokens | $2.50 | $2.25 |
| Results tracked | 20 | 51 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K2.5 leads
Grok 4.20 Multi-Agent: 43.0 (#92), Kimi K2.5: 48.8 (#53)
| Benchmark | Grok 4.20 Multi-Agent | Kimi K2.5 |
|---|---|---|
| LMArena Coding | 1457 | 1474 |
| SWE-bench Verified | — | 73.8% |
| SWE-bench Verified (bash only) | — | 70.8% |
| LMArena WebDev | — | 1437 |
| SWE-bench Multilingual | — | 67.3% |
| SciCode | — | 49% |
| WeirdML | — | 45.6% |
| ALE-Bench | — | 821.65 |
Agentic & Tool Use Not comparable
Grok 4.20 Multi-Agent: —, Kimi K2.5: 34.2 (#48)
| Benchmark | Grok 4.20 Multi-Agent | Kimi K2.5 |
|---|---|---|
| Terminal-Bench | — | 43.2% |
| OSWorld | — | 63.3% |
| LMArena Search | 1204 | — |
| Vending-Bench 2 | — | 1,198 |
Reasoning Grok 4.20 Multi-Agent leads
Grok 4.20 Multi-Agent: 43.9 (#48), Kimi K2.5: 31.2 (#80)
| Benchmark | Grok 4.20 Multi-Agent | Kimi K2.5 |
|---|---|---|
| NYT Connections (extended) | 89.6% | 69.9% |
| LMArena Hard Prompts | 1448 | 1453 |
| ARC-AGI-2 | — | 11.8% |
| SimpleBench | — | 46.8% |
| Kagi LLM Benchmark | — | 78.5% |
| ARC-AGI-1 | — | 65.3% |
| CritPt | — | 3.1% |
| Chess Puzzles | — | 12% |
| EnigmaEval | — | 3.4% |
| Thematic Generalization | — | 69.4% |
| Epoch Capabilities Index | — | 148.03 |
Math Kimi K2.5 leads
Grok 4.20 Multi-Agent: 39.4 (#104), Kimi K2.5: 51.8 (#53)
| Benchmark | Grok 4.20 Multi-Agent | Kimi K2.5 |
|---|---|---|
| LMArena Math | 1442 | 1470 |
| MathArena Final-Answer Competitions | — | 62.3% |
| OTIS Mock AIME 2024-2025 | — | 92.2% |
| FrontierMath (Feb 2025 set) | — | 27.9% |
| FrontierMath Tier 4 (v1) | — | 4.2% |
Knowledge Kimi K2.5 leads
Grok 4.20 Multi-Agent: 40.4 (#119), Kimi K2.5: 53.6 (#56)
| Benchmark | Grok 4.20 Multi-Agent | Kimi K2.5 |
|---|---|---|
| LMArena Expert | 1445 | 1466 |
| GPQA Diamond | — | 87.6% |
| Humanity's Last Exam | — | 24.4% |
| SimpleQA Verified | — | 34.3% |
| Vectara Hallucination Rate | — | 14.2% |
Multimodal Too close to call
Grok 4.20 Multi-Agent: 40.5 (#48), Kimi K2.5: 41.1 (#39)
| Benchmark | Grok 4.20 Multi-Agent | Kimi K2.5 |
|---|---|---|
| LMArena Vision | 1259 | 1269 |
| LMArena Document | — | 1430 |
Multilingual Too close to call
Grok 4.20 Multi-Agent: 54.4 (#43), Kimi K2.5: 53.9 (#53)
| Benchmark | Grok 4.20 Multi-Agent | Kimi K2.5 |
|---|---|---|
| LMArena Non-English | 1440 | 1433 |
| LMArena Chinese | 1475 | 1495 |
| LMArena French | 1466 | 1454 |
| LMArena German | 1456 | 1441 |
| LMArena Japanese | 1405 | 1421 |
| LMArena Korean | 1416 | 1410 |
| LMArena Russian | 1457 | 1435 |
| LMArena Spanish | 1447 | 1450 |
Instruction Following Too close to call
Grok 4.20 Multi-Agent: 74.8 (#84), Kimi K2.5: 75.3 (#64)
| Benchmark | Grok 4.20 Multi-Agent | Kimi K2.5 |
|---|---|---|
| LMArena Instruction Following | 1420 | 1431 |
Long Context Kimi K2.5 leads
Grok 4.20 Multi-Agent: 43.7 (#88), Kimi K2.5: 52.1 (#7)
| Benchmark | Grok 4.20 Multi-Agent | Kimi K2.5 |
|---|---|---|
| LMArena Longer Query | 1431 | 1445 |
| Fiction.LiveBench | — | 86.1% |
| CL-bench | — | 19.3% |
| CL-bench Life | — | 13.2% |
Writing & Preference Kimi K2.5 leads
Grok 4.20 Multi-Agent: 64.0 (#59), Kimi K2.5: 65.1 (#53)
| Benchmark | Grok 4.20 Multi-Agent | Kimi K2.5 |
|---|---|---|
| LMArena Text | 1450 | 1445 |
| LMArena Creative Writing | 1436 | 1423 |
| LMArena Multi-Turn | 1452 | 1444 |
| EQ-Bench Creative Writing | — | 1579 |
Frequently asked questions
Is Grok 4.20 Multi-Agent better than Kimi K2.5?
Kimi K2.5 is the stronger model overall, scoring 48.1 to 46.2 on the Noometry Index.
Which is cheaper, Grok 4.20 Multi-Agent or Kimi K2.5?
Kimi K2.5 is cheaper. It lists at $0.45 per million input tokens and $2.25 per million output tokens; Grok 4.20 Multi-Agent lists at $1.25 and $2.50.
Is Grok 4.20 Multi-Agent or Kimi K2.5 better for coding?
Kimi K2.5 scores higher on coding benchmarks: 48.8 versus 43.0 in the Noometry coding category.
Which has the bigger context window?
Grok 4.20 Multi-Agent does, with 1M tokens against 262K.
How many benchmarks do Grok 4.20 Multi-Agent and Kimi K2.5 share?
19 benchmarks have published results for both models. Grok 4.20 Multi-Agent has 20 scored results on Noometry and Kimi K2.5 has 51.