Model comparison
Kimi K2.6 vs Phi-4
Kimi K2.6 is the stronger model overall, scoring 47.7 to 31.2 on the Noometry Index. Phi-4 costs 20× less per token, which makes it the better buy when Kimi K2.6's lead doesn't matter for your workload.
Last verified . 22 shared benchmarks.
MicrosoftSummary
- They share 22 benchmarks with published results for both. Kimi K2.6 scores higher in 8 categories and Phi-4 in 1 category; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where Kimi K2.6 leads 57.0 to 20.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 96.1% for Kimi K2.6 and 13.8% for Phi-4.
- Phi-4 is cheaper at $0.07 / $0.14 per million input/output tokens, against $0.95 / $4 for Kimi K2.6.
- Kimi K2.6 accepts more context: 262K tokens versus 128K.
Side by side
| Kimi K2.6 | Phi-4 | |
|---|---|---|
| Provider | Moonshot AI | Microsoft |
| Noometry Index | 47.7 | 31.2 |
| Released | 2026-04-20 | 2024-12-11 |
| Weights | Open | Open |
| Context window | 262K | 128K |
| Max output | 262K | 4K |
| Input $ / M tokens | $0.95 | $0.07 |
| Output $ / M tokens | $4 | $0.14 |
| Results tracked | 51 | 37 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K2.6 leads
Kimi K2.6: 50.7 (#43), Phi-4: 34.4 (#239)
| Benchmark | Kimi K2.6 | Phi-4 |
|---|---|---|
| LMArena Coding | 1488 | 1231 |
| SWE-bench Verified | 76.7% | — |
| LMArena WebDev | 1509 | — |
| SciCode | 53.5% | — |
| WeirdML | 55.9% | — |
| BigCodeBench Instruct | — | 45.5% |
| LiveBench Coding | — | 30.7% |
| BigCodeBench Complete | — | 55.4% |
| ALE-Bench | 1,093 | — |
Agentic & Tool Use Too close to call
Kimi K2.6: 21.9 (#137), Phi-4: 22.8 (#128)
| Benchmark | Kimi K2.6 | Phi-4 |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 28.8% |
| OSWorld 2.0 | 4.6% | — |
| BALROG | — | 11.6% |
| ExploitBench | 18.4% | — |
| GBAEval | 0.9% | — |
| GDP.pdf | 12% | — |
| Vending-Bench 2 | 6,205 | — |
Reasoning Kimi K2.6 leads
Kimi K2.6: 40.5 (#55), Phi-4: 17.7 (#291)
| Benchmark | Kimi K2.6 | Phi-4 |
|---|---|---|
| Chess Puzzles | 26% | 1% |
| LMArena Hard Prompts | 1470 | 1220 |
| Epoch Capabilities Index | 151.05 | 130.42 |
| NYT Connections (extended) | 87.2% | — |
| CritPt | 8% | — |
| EBR-Bench | 2.4% | — |
| LiveBench Reasoning | — | 47.8% |
| Mystery Game Puzzles | 18% | — |
| DTBench | 90.9% | — |
| LiveBench Data Analysis | — | 45.2% |
| LMCA | 37.3% | — |
| LiveBench | — | 41.6% |
Math Kimi K2.6 leads
Kimi K2.6: 57.0 (#41), Phi-4: 20.8 (#285)
| Benchmark | Kimi K2.6 | Phi-4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 96.1% | 13.8% |
| LMArena Math | 1475 | 1246 |
| FrontierMath (Tiers 1-3) | 57.2% | — |
| FrontierMath Tier 4 | 25.6% | — |
| MathArena Final-Answer Competitions | 72.9% | — |
| ProofBench | 16% | — |
| LiveBench Math | — | 42% |
| MATH Level 5 | — | 64.9% |
| FrontierMath (Feb 2025 set) | 39% | — |
| FrontierMath Tier 4 (v1) | 14.6% | — |
Knowledge Kimi K2.6 leads
Kimi K2.6: 54.0 (#54), Phi-4: 32.6 (#209)
| Benchmark | Kimi K2.6 | Phi-4 |
|---|---|---|
| GPQA Diamond | 90.8% | 56.1% |
| Vectara Hallucination Rate | 10.8% | 3.7% |
| LMArena Expert | 1491 | 1203 |
| SimpleQA Verified | 34.9% | — |
| Confabulations | — | 29.4% |
| MMLU | — | 84.8% |
Multimodal Not comparable
Kimi K2.6: 31.6 (#103), Phi-4: —
| Benchmark | Kimi K2.6 | Phi-4 |
|---|---|---|
| LMArena Vision | 1283 | — |
| Blueprint-Bench 2 | 3.9% | — |
| Furniture Assembly | 21.7% | — |
| LMArena Document | 1451 | — |
Multilingual Kimi K2.6 leads
Kimi K2.6: 54.9 (#37), Phi-4: 37.2 (#237)
| Benchmark | Kimi K2.6 | Phi-4 |
|---|---|---|
| LMArena Non-English | 1446 | 1197 |
| LMArena Chinese | 1521 | 1212 |
| LMArena French | 1471 | 1224 |
| LMArena German | 1450 | 1222 |
| LMArena Japanese | 1443 | 1158 |
| LMArena Korean | 1427 | 1151 |
| LMArena Russian | 1446 | 1209 |
| LMArena Spanish | 1464 | 1234 |
Instruction Following Kimi K2.6 leads
Kimi K2.6: 76.3 (#43), Phi-4: 60.4 (#251)
| Benchmark | Kimi K2.6 | Phi-4 |
|---|---|---|
| LMArena Instruction Following | 1451 | 1201 |
| LiveBench Instruction Following | — | 58.4% |
Long Context Kimi K2.6 leads
Kimi K2.6: 44.9 (#52), Phi-4: 36.9 (#226)
| Benchmark | Kimi K2.6 | Phi-4 |
|---|---|---|
| LMArena Longer Query | 1468 | 1217 |
Writing & Preference Kimi K2.6 leads
Kimi K2.6: 68.5 (#26), Phi-4: 40.5 (#244)
| Benchmark | Kimi K2.6 | Phi-4 |
|---|---|---|
| LMArena Text | 1455 | 1217 |
| LMArena Creative Writing | 1434 | 1182 |
| LMArena Multi-Turn | 1453 | 1206 |
| Short-Story Creative Writing | — | 62.6% |
| EQ-Bench Creative Writing | 1725 | — |
| EQ-Bench 4 | 1202 | — |
| LiveBench Language | — | 25.6% |
Frequently asked questions
Is Kimi K2.6 better than Phi-4?
Kimi K2.6 is the stronger model overall, scoring 47.7 to 31.2 on the Noometry Index. Phi-4 costs 20× less per token, which makes it the better buy when Kimi K2.6's lead doesn't matter for your workload.
Which is cheaper, Kimi K2.6 or Phi-4?
Phi-4 is cheaper. It lists at $0.07 per million input tokens and $0.14 per million output tokens; Kimi K2.6 lists at $0.95 and $4.
Is Kimi K2.6 or Phi-4 better for coding?
Kimi K2.6 scores higher on coding benchmarks: 50.7 versus 34.4 in the Noometry coding category.
Which has the bigger context window?
Kimi K2.6 does, with 262K tokens against 128K.
How many benchmarks do Kimi K2.6 and Phi-4 share?
22 benchmarks have published results for both models. Kimi K2.6 has 51 scored results on Noometry and Phi-4 has 37.