Model comparison
Mistral Large 3 vs Phi-4
Mistral Large 3 is the stronger model overall, scoring 39.1 to 31.2 on the Noometry Index. Phi-4 costs 4.3× less per token, which makes it the better buy when Mistral Large 3's lead doesn't matter for your workload.
Last verified . 18 shared benchmarks.
Mistral AI
MicrosoftSummary
- They share 18 benchmarks with published results for both. Mistral Large 3 scores higher in 7 categories and Phi-4 in 1 category; 7 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where Mistral Large 3 leads 60.0 to 40.5.
- The biggest single-benchmark swing is Vectara Hallucination Rate: 14.5% for Mistral Large 3 and 3.7% for Phi-4.
- Phi-4 is cheaper at $0.07 / $0.14 per million input/output tokens, against $0.25 / $0.75 for Mistral Large 3.
- Mistral Large 3 accepts more context: 262K tokens versus 128K.
Side by side
| Mistral Large 3 | Phi-4 | |
|---|---|---|
| Provider | Mistral AI | Microsoft |
| Noometry Index | 39.1 | 31.2 |
| Released | 2025-12-02 | 2024-12-11 |
| Weights | Open | Open |
| Context window | 262K | 128K |
| Max output | 8K | 4K |
| Input $ / M tokens | $0.25 | $0.07 |
| Output $ / M tokens | $0.75 | $0.14 |
| Results tracked | 24 | 37 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
Mistral Large 3: 34.4 (#237), Phi-4: 34.4 (#239)
| Benchmark | Mistral Large 3 | Phi-4 |
|---|---|---|
| LMArena Coding | 1448 | 1231 |
| LMArena WebDev | 1230 | — |
| BigCodeBench Instruct | — | 45.5% |
| LiveBench Coding | — | 30.7% |
| BigCodeBench Complete | — | 55.4% |
Agentic & Tool Use Not comparable
Mistral Large 3: —, Phi-4: 22.8 (#128)
| Benchmark | Mistral Large 3 | Phi-4 |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 28.8% |
| BALROG | — | 11.6% |
Reasoning Phi-4 leads
Mistral Large 3: 15.2 (#319), Phi-4: 17.7 (#291)
| Benchmark | Mistral Large 3 | Phi-4 |
|---|---|---|
| LMArena Hard Prompts | 1429 | 1220 |
| Kagi LLM Benchmark | 50.9% | — |
| NYT Connections (extended) | 7.5% | — |
| Chess Puzzles | — | 1% |
| Thematic Generalization | 23% | — |
| LiveBench Reasoning | — | 47.8% |
| LiveBench Data Analysis | — | 45.2% |
| Epoch Capabilities Index | — | 130.42 |
| LiveBench | — | 41.6% |
Math Mistral Large 3 leads
Mistral Large 3: 38.7 (#129), Phi-4: 20.8 (#285)
| Benchmark | Mistral Large 3 | Phi-4 |
|---|---|---|
| LMArena Math | 1414 | 1246 |
| OTIS Mock AIME 2024-2025 | — | 13.8% |
| LiveBench Math | — | 42% |
| MATH Level 5 | — | 64.9% |
Knowledge Mistral Large 3 leads
Mistral Large 3: 36.0 (#177), Phi-4: 32.6 (#209)
| Benchmark | Mistral Large 3 | Phi-4 |
|---|---|---|
| Vectara Hallucination Rate | 14.5% | 3.7% |
| LMArena Expert | 1421 | 1203 |
| GPQA Diamond | — | 56.1% |
| Confabulations | — | 29.4% |
| MMLU | — | 84.8% |
Multimodal Not comparable
Mistral Large 3: 38.2 (#66), Phi-4: —
| Benchmark | Mistral Large 3 | Phi-4 |
|---|---|---|
| LMArena Vision | 1221 | — |
Multilingual Mistral Large 3 leads
Mistral Large 3: 52.5 (#84), Phi-4: 37.2 (#237)
| Benchmark | Mistral Large 3 | Phi-4 |
|---|---|---|
| LMArena Non-English | 1413 | 1197 |
| LMArena Chinese | 1447 | 1212 |
| LMArena French | 1455 | 1224 |
| LMArena German | 1437 | 1222 |
| LMArena Japanese | 1394 | 1158 |
| LMArena Korean | 1384 | 1151 |
| LMArena Russian | 1411 | 1209 |
| LMArena Spanish | 1440 | 1234 |
Instruction Following Mistral Large 3 leads
Mistral Large 3: 74.0 (#108), Phi-4: 60.4 (#251)
| Benchmark | Mistral Large 3 | Phi-4 |
|---|---|---|
| LMArena Instruction Following | 1403 | 1201 |
| LiveBench Instruction Following | — | 58.4% |
Long Context Mistral Large 3 leads
Mistral Large 3: 43.1 (#105), Phi-4: 36.9 (#226)
| Benchmark | Mistral Large 3 | Phi-4 |
|---|---|---|
| LMArena Longer Query | 1413 | 1217 |
Writing & Preference Mistral Large 3 leads
Mistral Large 3: 60.0 (#101), Phi-4: 40.5 (#244)
| Benchmark | Mistral Large 3 | Phi-4 |
|---|---|---|
| LMArena Text | 1428 | 1217 |
| LMArena Creative Writing | 1386 | 1182 |
| LMArena Multi-Turn | 1429 | 1206 |
| Short-Story Creative Writing | — | 62.6% |
| EQ-Bench Creative Writing | 1412 | — |
| LiveBench Language | — | 25.6% |
Frequently asked questions
Is Mistral Large 3 better than Phi-4?
Mistral Large 3 is the stronger model overall, scoring 39.1 to 31.2 on the Noometry Index. Phi-4 costs 4.3× less per token, which makes it the better buy when Mistral Large 3's lead doesn't matter for your workload.
Which is cheaper, Mistral Large 3 or Phi-4?
Phi-4 is cheaper. It lists at $0.07 per million input tokens and $0.14 per million output tokens; Mistral Large 3 lists at $0.25 and $0.75.
Is Mistral Large 3 or Phi-4 better for coding?
They score almost the same on coding (34.4 vs 34.4); test both on your own repository before choosing.
Which has the bigger context window?
Mistral Large 3 does, with 262K tokens against 128K.
How many benchmarks do Mistral Large 3 and Phi-4 share?
18 benchmarks have published results for both models. Mistral Large 3 has 24 scored results on Noometry and Phi-4 has 37.