Model comparison
GLM-4.6 vs Phi-4
GLM-4.6 is the stronger model overall, scoring 41.4 to 31.2 on the Noometry Index. Phi-4 costs 11× less per token, which makes it the better buy when GLM-4.6's lead doesn't matter for your workload.
Last verified . 19 shared benchmarks.
MicrosoftSummary
- They share 19 benchmarks with published results for both. GLM-4.6 scores higher in 9 categories and Phi-4 in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where GLM-4.6 leads 61.1 to 40.5.
- The biggest single-benchmark swing is Berkeley Function Calling Leaderboard: 72.4% for GLM-4.6 and 28.8% for Phi-4.
- Phi-4 is cheaper at $0.07 / $0.14 per million input/output tokens, against $0.60 / $2.20 for GLM-4.6.
- GLM-4.6 accepts more context: 205K tokens versus 128K.
Side by side
| GLM-4.6 | Phi-4 | |
|---|---|---|
| Provider | Z.ai (Zhipu) | Microsoft |
| Noometry Index | 41.4 | 31.2 |
| Released | 2025-09-30 | 2024-12-11 |
| Weights | Open | Open |
| Context window | 205K | 128K |
| Max output | 131K | 4K |
| Input $ / M tokens | $0.60 | $0.07 |
| Output $ / M tokens | $2.20 | $0.14 |
| Results tracked | 29 | 37 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GLM-4.6 leads
GLM-4.6: 40.1 (#148), Phi-4: 34.4 (#239)
| Benchmark | GLM-4.6 | Phi-4 |
|---|---|---|
| LMArena Coding | 1449 | 1231 |
| SWE-bench Verified (bash only) | 55.4% | — |
| LMArena WebDev | 1340 | — |
| SciCode | 38.4% | — |
| BigCodeBench Instruct | — | 45.5% |
| LiveBench Coding | — | 30.7% |
| BigCodeBench Complete | — | 55.4% |
| ALE-Bench | 340.82 | — |
Agentic & Tool Use GLM-4.6 leads
GLM-4.6: 32.3 (#66), Phi-4: 22.8 (#128)
| Benchmark | GLM-4.6 | Phi-4 |
|---|---|---|
| Berkeley Function Calling Leaderboard | 72.4% | 28.8% |
| Terminal-Bench | 24.5% | — |
| BALROG | — | 11.6% |
Reasoning GLM-4.6 leads
GLM-4.6: 23.7 (#172), Phi-4: 17.7 (#291)
| Benchmark | GLM-4.6 | Phi-4 |
|---|---|---|
| LMArena Hard Prompts | 1440 | 1220 |
| Kagi LLM Benchmark | 47.4% | — |
| CritPt | 1.1% | — |
| Chess Puzzles | — | 1% |
| LiveBench Reasoning | — | 47.8% |
| LiveBench Data Analysis | — | 45.2% |
| Epoch Capabilities Index | — | 130.42 |
| LiveBench | — | 41.6% |
Math GLM-4.6 leads
GLM-4.6: 39.1 (#111), Phi-4: 20.8 (#285)
| Benchmark | GLM-4.6 | Phi-4 |
|---|---|---|
| LMArena Math | 1432 | 1246 |
| OTIS Mock AIME 2024-2025 | — | 13.8% |
| LiveBench Math | — | 42% |
| MATH Level 5 | — | 64.9% |
| FrontierMath (Feb 2025 set) | 3.8% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge GLM-4.6 leads
GLM-4.6: 40.2 (#124), Phi-4: 32.6 (#209)
| Benchmark | GLM-4.6 | Phi-4 |
|---|---|---|
| Vectara Hallucination Rate | 9.5% | 3.7% |
| LMArena Expert | 1431 | 1203 |
| GPQA Diamond | — | 56.1% |
| Confabulations | — | 29.4% |
| MMLU | — | 84.8% |
Multilingual GLM-4.6 leads
GLM-4.6: 53.5 (#66), Phi-4: 37.2 (#237)
| Benchmark | GLM-4.6 | Phi-4 |
|---|---|---|
| LMArena Non-English | 1426 | 1197 |
| LMArena Chinese | 1499 | 1212 |
| LMArena French | 1459 | 1224 |
| LMArena German | 1447 | 1222 |
| LMArena Japanese | 1393 | 1158 |
| LMArena Korean | 1400 | 1151 |
| LMArena Russian | 1419 | 1209 |
| LMArena Spanish | 1436 | 1234 |
Instruction Following GLM-4.6 leads
GLM-4.6: 74.3 (#98), Phi-4: 60.4 (#251)
| Benchmark | GLM-4.6 | Phi-4 |
|---|---|---|
| LMArena Instruction Following | 1410 | 1201 |
| LiveBench Instruction Following | — | 58.4% |
Long Context GLM-4.6 leads
GLM-4.6: 43.4 (#94), Phi-4: 36.9 (#226)
| Benchmark | GLM-4.6 | Phi-4 |
|---|---|---|
| LMArena Longer Query | 1422 | 1217 |
Writing & Preference GLM-4.6 leads
GLM-4.6: 61.1 (#90), Phi-4: 40.5 (#244)
| Benchmark | GLM-4.6 | Phi-4 |
|---|---|---|
| LMArena Text | 1440 | 1217 |
| LMArena Creative Writing | 1411 | 1182 |
| LMArena Multi-Turn | 1427 | 1206 |
| Short-Story Creative Writing | — | 62.6% |
| EQ-Bench Creative Writing | 1411 | — |
| LiveBench Language | — | 25.6% |
Frequently asked questions
Is GLM-4.6 better than Phi-4?
GLM-4.6 is the stronger model overall, scoring 41.4 to 31.2 on the Noometry Index. Phi-4 costs 11× less per token, which makes it the better buy when GLM-4.6's lead doesn't matter for your workload.
Which is cheaper, GLM-4.6 or Phi-4?
Phi-4 is cheaper. It lists at $0.07 per million input tokens and $0.14 per million output tokens; GLM-4.6 lists at $0.60 and $2.20.
Is GLM-4.6 or Phi-4 better for coding?
GLM-4.6 scores higher on coding benchmarks: 40.1 versus 34.4 in the Noometry coding category.
Which has the bigger context window?
GLM-4.6 does, with 205K tokens against 128K.
How many benchmarks do GLM-4.6 and Phi-4 share?
19 benchmarks have published results for both models. GLM-4.6 has 29 scored results on Noometry and Phi-4 has 37.