Model comparison
Phi-4 vs Qwen3.5 35B-A3B
Qwen3.5 35B-A3B is the stronger model overall, scoring 42.0 to 31.2 on the Noometry Index. Phi-4 costs 7.9× less per token, which makes it the better buy when Qwen3.5 35B-A3B's lead doesn't matter for your workload.
Last verified . 22 shared benchmarks.
Microsoft
Alibaba (Qwen)Summary
- They share 22 benchmarks with published results for both. Phi-4 scores higher in 1 category and Qwen3.5 35B-A3B in 7 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in math, where Qwen3.5 35B-A3B leads 39.9 to 20.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 13.8% for Phi-4 and 70% for Qwen3.5 35B-A3B.
- Phi-4 is cheaper at $0.07 / $0.14 per million input/output tokens, against $0.25 / $2 for Qwen3.5 35B-A3B.
- Qwen3.5 35B-A3B accepts more context: 262K tokens versus 128K.
Side by side
| Phi-4 | Qwen3.5 35B-A3B | |
|---|---|---|
| Provider | Microsoft | Alibaba (Qwen) |
| Noometry Index | 31.2 | 42.0 |
| Released | 2024-12-11 | 2026-02-01 |
| Weights | Open | Open |
| Context window | 128K | 262K |
| Max output | 4K | 66K |
| Input $ / M tokens | $0.07 | $0.25 |
| Output $ / M tokens | $0.14 | $2 |
| Results tracked | 37 | 28 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
Phi-4: 34.4 (#239), Qwen3.5 35B-A3B: 33.8 (#251)
| Benchmark | Phi-4 | Qwen3.5 35B-A3B |
|---|---|---|
| LMArena Coding | 1231 | 1410 |
| LMArena WebDev | — | 1254 |
| SciCode | — | 29.3% |
| BigCodeBench Instruct | 45.5% | — |
| LiveBench Coding | 30.7% | — |
| BigCodeBench Complete | 55.4% | — |
Agentic & Tool Use Not comparable
Phi-4: 22.8 (#128), Qwen3.5 35B-A3B: —
| Benchmark | Phi-4 | Qwen3.5 35B-A3B |
|---|---|---|
| Berkeley Function Calling Leaderboard | 28.8% | — |
| BALROG | 11.6% | — |
Reasoning Qwen3.5 35B-A3B leads
Phi-4: 17.7 (#291), Qwen3.5 35B-A3B: 24.6 (#161)
| Benchmark | Phi-4 | Qwen3.5 35B-A3B |
|---|---|---|
| Chess Puzzles | 1% | 10% |
| LMArena Hard Prompts | 1220 | 1400 |
| Epoch Capabilities Index | 130.42 | 142.52 |
| CritPt | — | 0.6% |
| LiveBench Reasoning | 47.8% | — |
| DTBench | — | 80% |
| LiveBench Data Analysis | 45.2% | — |
| LMCA | — | 29.5% |
| LiveBench | 41.6% | — |
Math Qwen3.5 35B-A3B leads
Phi-4: 20.8 (#285), Qwen3.5 35B-A3B: 39.9 (#97)
| Benchmark | Phi-4 | Qwen3.5 35B-A3B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 13.8% | 70% |
| LMArena Math | 1246 | 1404 |
| MathArena Final-Answer Competitions | — | 56% |
| LiveBench Math | 42% | — |
| MATH Level 5 | 64.9% | — |
Knowledge Qwen3.5 35B-A3B leads
Phi-4: 32.6 (#209), Qwen3.5 35B-A3B: 47.8 (#79)
| Benchmark | Phi-4 | Qwen3.5 35B-A3B |
|---|---|---|
| GPQA Diamond | 56.1% | 83.5% |
| Vectara Hallucination Rate | 3.7% | 10.5% |
| LMArena Expert | 1203 | 1408 |
| Confabulations | 29.4% | — |
| MMLU | 84.8% | — |
Multilingual Qwen3.5 35B-A3B leads
Phi-4: 37.2 (#237), Qwen3.5 35B-A3B: 50.0 (#127)
| Benchmark | Phi-4 | Qwen3.5 35B-A3B |
|---|---|---|
| LMArena Non-English | 1197 | 1378 |
| LMArena Chinese | 1212 | 1457 |
| LMArena French | 1224 | 1412 |
| LMArena German | 1222 | 1367 |
| LMArena Japanese | 1158 | 1325 |
| LMArena Korean | 1151 | 1356 |
| LMArena Russian | 1209 | 1376 |
| LMArena Spanish | 1234 | 1392 |
Instruction Following Qwen3.5 35B-A3B leads
Phi-4: 60.4 (#251), Qwen3.5 35B-A3B: 72.8 (#128)
| Benchmark | Phi-4 | Qwen3.5 35B-A3B |
|---|---|---|
| LMArena Instruction Following | 1201 | 1379 |
| LiveBench Instruction Following | 58.4% | — |
Long Context Qwen3.5 35B-A3B leads
Phi-4: 36.9 (#226), Qwen3.5 35B-A3B: 42.4 (#127)
| Benchmark | Phi-4 | Qwen3.5 35B-A3B |
|---|---|---|
| LMArena Longer Query | 1217 | 1389 |
Writing & Preference Qwen3.5 35B-A3B leads
Phi-4: 40.5 (#244), Qwen3.5 35B-A3B: 57.9 (#124)
| Benchmark | Phi-4 | Qwen3.5 35B-A3B |
|---|---|---|
| LMArena Text | 1217 | 1395 |
| LMArena Creative Writing | 1182 | 1346 |
| LMArena Multi-Turn | 1206 | 1390 |
| Short-Story Creative Writing | 62.6% | — |
| LiveBench Language | 25.6% | — |
Frequently asked questions
Is Phi-4 better than Qwen3.5 35B-A3B?
Qwen3.5 35B-A3B is the stronger model overall, scoring 42.0 to 31.2 on the Noometry Index. Phi-4 costs 7.9× less per token, which makes it the better buy when Qwen3.5 35B-A3B's lead doesn't matter for your workload.
Which is cheaper, Phi-4 or Qwen3.5 35B-A3B?
Phi-4 is cheaper. It lists at $0.07 per million input tokens and $0.14 per million output tokens; Qwen3.5 35B-A3B lists at $0.25 and $2.
Is Phi-4 or Qwen3.5 35B-A3B better for coding?
They score almost the same on coding (34.4 vs 33.8); test both on your own repository before choosing.
Which has the bigger context window?
Qwen3.5 35B-A3B does, with 262K tokens against 128K.
How many benchmarks do Phi-4 and Qwen3.5 35B-A3B share?
22 benchmarks have published results for both models. Phi-4 has 37 scored results on Noometry and Qwen3.5 35B-A3B has 28.