Model comparison
Phi-4 vs Qwen3-30B-A3B
Qwen3-30B-A3B is the stronger model overall, scoring 38.9 to 31.2 on the Noometry Index. Phi-4 costs 2.5× less per token, which makes it the better buy when Qwen3-30B-A3B's lead doesn't matter for your workload.
Last verified . 24 shared benchmarks.
Microsoft
Alibaba (Qwen)Summary
- They share 24 benchmarks with published results for both. Phi-4 scores higher in 1 category and Qwen3-30B-A3B in 8 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where Qwen3-30B-A3B leads 37.4 to 20.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 13.8% for Phi-4 and 70.3% for Qwen3-30B-A3B.
- Phi-4 is cheaper at $0.07 / $0.14 per million input/output tokens, against $0.12 / $0.50 for Qwen3-30B-A3B.
- Phi-4 accepts more context: 128K tokens versus 41K.
Side by side
| Phi-4 | Qwen3-30B-A3B | |
|---|---|---|
| Provider | Microsoft | Alibaba (Qwen) |
| Noometry Index | 31.2 | 38.9 |
| Released | 2024-12-11 | 2025-04-28 |
| Weights | Open | Open |
| Context window | 128K | 41K |
| Max output | 4K | 16K |
| Input $ / M tokens | $0.07 | $0.12 |
| Output $ / M tokens | $0.14 | $0.50 |
| Results tracked | 37 | 32 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3-30B-A3B leads
Phi-4: 34.4 (#239), Qwen3-30B-A3B: 37.5 (#194)
| Benchmark | Phi-4 | Qwen3-30B-A3B |
|---|---|---|
| LMArena Coding | 1231 | 1416 |
| SciCode | — | 33.3% |
| WeirdML | — | 29.8% |
| BigCodeBench Instruct | 45.5% | — |
| LiveBench Coding | 30.7% | — |
| BigCodeBench Complete | 55.4% | — |
Agentic & Tool Use Qwen3-30B-A3B leads
Phi-4: 22.8 (#128), Qwen3-30B-A3B: 29.8 (#82)
| Benchmark | Phi-4 | Qwen3-30B-A3B |
|---|---|---|
| Berkeley Function Calling Leaderboard | 28.8% | 41.4% |
| BALROG | 11.6% | — |
Reasoning Qwen3-30B-A3B leads
Phi-4: 17.7 (#291), Qwen3-30B-A3B: 22.2 (#204)
| Benchmark | Phi-4 | Qwen3-30B-A3B |
|---|---|---|
| Chess Puzzles | 1% | 8% |
| LMArena Hard Prompts | 1220 | 1398 |
| Epoch Capabilities Index | 130.42 | 139.63 |
| Kagi LLM Benchmark | — | 54.9% |
| CritPt | — | 0.3% |
| LiveBench Reasoning | 47.8% | — |
| DTBench | — | 69.3% |
| LiveBench Data Analysis | 45.2% | — |
| LMCA | — | 22.4% |
| LiveBench | 41.6% | — |
Math Qwen3-30B-A3B leads
Phi-4: 20.8 (#285), Qwen3-30B-A3B: 37.4 (#157)
| Benchmark | Phi-4 | Qwen3-30B-A3B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 13.8% | 70.3% |
| LMArena Math | 1246 | 1394 |
| MathArena Final-Answer Competitions | — | 47.8% |
| LiveBench Math | 42% | — |
| MATH Level 5 | 64.9% | — |
Knowledge Qwen3-30B-A3B leads
Phi-4: 32.6 (#209), Qwen3-30B-A3B: 41.8 (#105)
| Benchmark | Phi-4 | Qwen3-30B-A3B |
|---|---|---|
| GPQA Diamond | 56.1% | 70.1% |
| Confabulations | 29.4% | 12.3% |
| LMArena Expert | 1203 | 1396 |
| Vectara Hallucination Rate | 3.7% | — |
| MMLU | 84.8% | — |
Multilingual Qwen3-30B-A3B leads
Phi-4: 37.2 (#237), Qwen3-30B-A3B: 49.5 (#132)
| Benchmark | Phi-4 | Qwen3-30B-A3B |
|---|---|---|
| LMArena Non-English | 1197 | 1372 |
| LMArena Chinese | 1212 | 1433 |
| LMArena French | 1224 | 1418 |
| LMArena German | 1222 | 1380 |
| LMArena Japanese | 1158 | 1337 |
| LMArena Korean | 1151 | 1331 |
| LMArena Russian | 1209 | 1370 |
| LMArena Spanish | 1234 | 1404 |
Instruction Following Qwen3-30B-A3B leads
Phi-4: 60.4 (#251), Qwen3-30B-A3B: 72.0 (#142)
| Benchmark | Phi-4 | Qwen3-30B-A3B |
|---|---|---|
| LMArena Instruction Following | 1201 | 1363 |
| LiveBench Instruction Following | 58.4% | — |
Long Context Phi-4 leads
Phi-4: 36.9 (#226), Qwen3-30B-A3B: 31.0 (#283)
| Benchmark | Phi-4 | Qwen3-30B-A3B |
|---|---|---|
| LMArena Longer Query | 1217 | 1379 |
| Fiction.LiveBench | — | 40.6% |
Writing & Preference Qwen3-30B-A3B leads
Phi-4: 40.5 (#244), Qwen3-30B-A3B: 55.6 (#143)
| Benchmark | Phi-4 | Qwen3-30B-A3B |
|---|---|---|
| LMArena Text | 1217 | 1384 |
| LMArena Creative Writing | 1182 | 1317 |
| Short-Story Creative Writing | 62.6% | 75.3% |
| LMArena Multi-Turn | 1206 | 1378 |
| LiveBench Language | 25.6% | — |
Frequently asked questions
Is Phi-4 better than Qwen3-30B-A3B?
Qwen3-30B-A3B is the stronger model overall, scoring 38.9 to 31.2 on the Noometry Index. Phi-4 costs 2.5× less per token, which makes it the better buy when Qwen3-30B-A3B's lead doesn't matter for your workload.
Which is cheaper, Phi-4 or Qwen3-30B-A3B?
Phi-4 is cheaper. It lists at $0.07 per million input tokens and $0.14 per million output tokens; Qwen3-30B-A3B lists at $0.12 and $0.50.
Is Phi-4 or Qwen3-30B-A3B better for coding?
Qwen3-30B-A3B scores higher on coding benchmarks: 37.5 versus 34.4 in the Noometry coding category.
Which has the bigger context window?
Phi-4 does, with 128K tokens against 41K.
How many benchmarks do Phi-4 and Qwen3-30B-A3B share?
24 benchmarks have published results for both models. Phi-4 has 37 scored results on Noometry and Qwen3-30B-A3B has 32.