Model comparison
MiniMax-M2.5 vs Phi-4
MiniMax-M2.5 is the stronger model overall, scoring 38.3 to 31.2 on the Noometry Index. Phi-4 costs 6.0× less per token, which makes it the better buy when MiniMax-M2.5's lead doesn't matter for your workload.
Last verified . 19 shared benchmarks.
MiniMax
MicrosoftSummary
- They share 19 benchmarks with published results for both. MiniMax-M2.5 scores higher in 8 categories and Phi-4 in 1 category; 7 gaps are clear of the uncertainty.
- The widest gap is in coding, where MiniMax-M2.5 leads 48.1 to 34.4.
- The biggest single-benchmark swing is Vectara Hallucination Rate: 9.1% for MiniMax-M2.5 and 3.7% for Phi-4.
- Phi-4 is cheaper at $0.07 / $0.14 per million input/output tokens, against $0.30 / $1.20 for MiniMax-M2.5.
- MiniMax-M2.5 accepts more context: 205K tokens versus 128K.
Side by side
| MiniMax-M2.5 | Phi-4 | |
|---|---|---|
| Provider | MiniMax | Microsoft |
| Noometry Index | 38.3 | 31.2 |
| Released | 2026-02-12 | 2024-12-11 |
| Weights | Open | Open |
| Context window | 205K | 128K |
| Max output | 131K | 4K |
| Input $ / M tokens | $0.30 | $0.07 |
| Output $ / M tokens | $1.20 | $0.14 |
| Results tracked | 33 | 37 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding MiniMax-M2.5 leads
MiniMax-M2.5: 48.1 (#58), Phi-4: 34.4 (#239)
| Benchmark | MiniMax-M2.5 | Phi-4 |
|---|---|---|
| LMArena Coding | 1381 | 1231 |
| SWE-bench Verified (bash only) | 75.8% | — |
| LMArena WebDev | 1387 | — |
| SWE-bench Multilingual | 68.3% | — |
| BigCodeBench Instruct | — | 45.5% |
| LiveBench Coding | — | 30.7% |
| BigCodeBench Complete | — | 55.4% |
| ALE-Bench | 618.17 | — |
Agentic & Tool Use MiniMax-M2.5 leads
MiniMax-M2.5: 30.4 (#77), Phi-4: 22.8 (#128)
| Benchmark | MiniMax-M2.5 | Phi-4 |
|---|---|---|
| Terminal-Bench | 42.7% | — |
| Berkeley Function Calling Leaderboard | — | 28.8% |
| BALROG | — | 11.6% |
| Vending-Bench 2 | -23.16 | — |
Reasoning Too close to call
MiniMax-M2.5: 17.5 (#292), Phi-4: 17.7 (#291)
| Benchmark | MiniMax-M2.5 | Phi-4 |
|---|---|---|
| LMArena Hard Prompts | 1372 | 1220 |
| Epoch Capabilities Index | 146.68 | 130.42 |
| ARC-AGI-2 | 4.9% | — |
| Kagi LLM Benchmark | 55.2% | — |
| NYT Connections (extended) | 16.8% | — |
| ARC-AGI-1 | 63.7% | — |
| Chess Puzzles | — | 1% |
| LiveBench Reasoning | — | 47.8% |
| LiveBench Data Analysis | — | 45.2% |
| LiveBench | — | 41.6% |
Math MiniMax-M2.5 leads
MiniMax-M2.5: 26.9 (#253), Phi-4: 20.8 (#285)
| Benchmark | MiniMax-M2.5 | Phi-4 |
|---|---|---|
| LMArena Math | 1378 | 1246 |
| OTIS Mock AIME 2024-2025 | — | 13.8% |
| ProofBench | 4% | — |
| LiveBench Math | — | 42% |
| MATH Level 5 | — | 64.9% |
Knowledge MiniMax-M2.5 leads
MiniMax-M2.5: 39.2 (#135), Phi-4: 32.6 (#209)
| Benchmark | MiniMax-M2.5 | Phi-4 |
|---|---|---|
| Vectara Hallucination Rate | 9.1% | 3.7% |
| LMArena Expert | 1379 | 1203 |
| GPQA Diamond | — | 56.1% |
| Confabulations | — | 29.4% |
| MMLU | — | 84.8% |
Multilingual MiniMax-M2.5 leads
MiniMax-M2.5: 47.1 (#152), Phi-4: 37.2 (#237)
| Benchmark | MiniMax-M2.5 | Phi-4 |
|---|---|---|
| LMArena Non-English | 1338 | 1197 |
| LMArena Chinese | 1393 | 1212 |
| LMArena French | 1362 | 1224 |
| LMArena German | 1362 | 1222 |
| LMArena Japanese | 1171 | 1158 |
| LMArena Korean | 1232 | 1151 |
| LMArena Russian | 1358 | 1209 |
| LMArena Spanish | 1354 | 1234 |
Instruction Following MiniMax-M2.5 leads
MiniMax-M2.5: 71.5 (#148), Phi-4: 60.4 (#251)
| Benchmark | MiniMax-M2.5 | Phi-4 |
|---|---|---|
| LMArena Instruction Following | 1353 | 1201 |
| LiveBench Instruction Following | — | 58.4% |
Long Context Too close to call
MiniMax-M2.5: 37.5 (#216), Phi-4: 36.9 (#226)
| Benchmark | MiniMax-M2.5 | Phi-4 |
|---|---|---|
| LMArena Longer Query | 1366 | 1217 |
| CL-bench | 11.4% | — |
| CL-bench Life | 6.3% | — |
Writing & Preference MiniMax-M2.5 leads
MiniMax-M2.5: 53.9 (#153), Phi-4: 40.5 (#244)
| Benchmark | MiniMax-M2.5 | Phi-4 |
|---|---|---|
| LMArena Text | 1359 | 1217 |
| LMArena Creative Writing | 1331 | 1182 |
| LMArena Multi-Turn | 1364 | 1206 |
| Short-Story Creative Writing | — | 62.6% |
| EQ-Bench Creative Writing | 1361 | — |
| LiveBench Language | — | 25.6% |
Frequently asked questions
Is MiniMax-M2.5 better than Phi-4?
MiniMax-M2.5 is the stronger model overall, scoring 38.3 to 31.2 on the Noometry Index. Phi-4 costs 6.0× less per token, which makes it the better buy when MiniMax-M2.5's lead doesn't matter for your workload.
Which is cheaper, MiniMax-M2.5 or Phi-4?
Phi-4 is cheaper. It lists at $0.07 per million input tokens and $0.14 per million output tokens; MiniMax-M2.5 lists at $0.30 and $1.20.
Is MiniMax-M2.5 or Phi-4 better for coding?
MiniMax-M2.5 scores higher on coding benchmarks: 48.1 versus 34.4 in the Noometry coding category.
Which has the bigger context window?
MiniMax-M2.5 does, with 205K tokens against 128K.
How many benchmarks do MiniMax-M2.5 and Phi-4 share?
19 benchmarks have published results for both models. MiniMax-M2.5 has 33 scored results on Noometry and Phi-4 has 37.