Model comparison
Phi-4 vs Trinity Large Thinking
Trinity Large Thinking is the stronger model overall, scoring 38.1 to 30.9 on the Noometry Index. Phi-4 costs 4.4× less per token, which makes it the better buy when Trinity Large Thinking's lead doesn't matter for your workload.
Last verified . 18 shared benchmarks.
Microsoft
Arcee AISummary
- They share 18 benchmarks with published results for both. Phi-4 scores higher in 2 categories and Trinity Large Thinking in 6 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in math, where Trinity Large Thinking leads 34.9 to 19.6.
- Phi-4 is cheaper at $0.07 / $0.14 per million input/output tokens, against $0.25 / $0.80 for Trinity Large Thinking.
- Trinity Large Thinking accepts more context: 262K tokens versus 128K.
Side by side
| Phi-4 | Trinity Large Thinking | |
|---|---|---|
| Provider | Microsoft | Arcee AI |
| Noometry Index | 30.9 | 38.1 |
| Released | 2024-12-11 | 2026-04-01 |
| Weights | Open | Open |
| Context window | 128K | 262K |
| Max output | 4K | 80K |
| Input $ / M tokens | $0.07 | $0.25 |
| Output $ / M tokens | $0.14 | $0.80 |
| Results tracked | 37 | 24 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
Phi-4: 33.8 (#237), Trinity Large Thinking: 33.8 (#238)
| Benchmark | Phi-4 | Trinity Large Thinking |
|---|---|---|
| LMArena Coding | 1231 | 1381 |
| LMArena WebDev | — | 1238 |
| SciCode | — | 36.1% |
| BigCodeBench Instruct | 45.5% | — |
| LiveBench Coding | 30.7% | — |
| BigCodeBench Complete | 55.4% | — |
Agentic & Tool Use Not comparable
Phi-4: 22.6 (#120), Trinity Large Thinking: —
| Benchmark | Phi-4 | Trinity Large Thinking |
|---|---|---|
| Berkeley Function Calling Leaderboard | 28.8% | — |
| BALROG | 11.6% | — |
Reasoning Too close to call
Phi-4: 17.0 (#287), Trinity Large Thinking: 16.3 (#298)
| Benchmark | Phi-4 | Trinity Large Thinking |
|---|---|---|
| LMArena Hard Prompts | 1220 | 1350 |
| NYT Connections (extended) | — | 16.5% |
| CritPt | — | 0.9% |
| Chess Puzzles | 1% | — |
| Thematic Generalization | — | 41.6% |
| LiveBench Reasoning | 47.8% | — |
| LiveBench Data Analysis | 45.2% | — |
| Surface Evolver Bench | — | 15.6% |
| Epoch Capabilities Index | 130.43 | — |
| LiveBench | 41.6% | — |
Math Trinity Large Thinking leads
Phi-4: 19.6 (#279), Trinity Large Thinking: 34.9 (#149)
| Benchmark | Phi-4 | Trinity Large Thinking |
|---|---|---|
| LMArena Math | 1246 | 1366 |
| OTIS Mock AIME 2024-2025 | 13.8% | — |
| LiveBench Math | 42% | — |
| MATH Level 5 | 64.9% | — |
Knowledge Trinity Large Thinking leads
Phi-4: 32.5 (#204), Trinity Large Thinking: 40.3 (#107)
| Benchmark | Phi-4 | Trinity Large Thinking |
|---|---|---|
| Vectara Hallucination Rate | 3.7% | 6.9% |
| LMArena Expert | 1203 | 1359 |
| GPQA Diamond | 56.1% | — |
| Confabulations | 29.4% | — |
| MMLU | 84.8% | — |
Multilingual Trinity Large Thinking leads
Phi-4: 37.4 (#232), Trinity Large Thinking: 46.4 (#155)
| Benchmark | Phi-4 | Trinity Large Thinking |
|---|---|---|
| LMArena Non-English | 1197 | 1325 |
| LMArena Chinese | 1211 | 1372 |
| LMArena French | 1223 | 1374 |
| LMArena German | 1222 | 1356 |
| LMArena Japanese | 1158 | 1312 |
| LMArena Korean | 1151 | 1304 |
| LMArena Russian | 1209 | 1338 |
| LMArena Spanish | 1234 | 1356 |
Instruction Following Trinity Large Thinking leads
Phi-4: 60.6 (#246), Trinity Large Thinking: 70.7 (#157)
| Benchmark | Phi-4 | Trinity Large Thinking |
|---|---|---|
| LMArena Instruction Following | 1201 | 1334 |
| LiveBench Instruction Following | 58.4% | — |
Long Context Trinity Large Thinking leads
Phi-4: 36.9 (#221), Trinity Large Thinking: 41.2 (#139)
| Benchmark | Phi-4 | Trinity Large Thinking |
|---|---|---|
| LMArena Longer Query | 1217 | 1355 |
Writing & Preference Trinity Large Thinking leads
Phi-4: 40.7 (#239), Trinity Large Thinking: 54.0 (#152)
| Benchmark | Phi-4 | Trinity Large Thinking |
|---|---|---|
| LMArena Text | 1217 | 1340 |
| LMArena Creative Writing | 1182 | 1321 |
| LMArena Multi-Turn | 1206 | 1342 |
| Short-Story Creative Writing | 62.6% | — |
| LiveBench Language | 25.6% | — |
Frequently asked questions
Is Phi-4 better than Trinity Large Thinking?
Trinity Large Thinking is the stronger model overall, scoring 38.1 to 30.9 on the Noometry Index. Phi-4 costs 4.4× less per token, which makes it the better buy when Trinity Large Thinking's lead doesn't matter for your workload.
Which is cheaper, Phi-4 or Trinity Large Thinking?
Phi-4 is cheaper. It lists at $0.07 per million input tokens and $0.14 per million output tokens; Trinity Large Thinking lists at $0.25 and $0.80.
Is Phi-4 or Trinity Large Thinking better for coding?
They score almost the same on coding (33.8 vs 33.8); test both on your own repository before choosing.
Which has the bigger context window?
Trinity Large Thinking does, with 262K tokens against 128K.
How many benchmarks do Phi-4 and Trinity Large Thinking share?
18 benchmarks have published results for both models. Phi-4 has 37 scored results on Noometry and Trinity Large Thinking has 24.