Model comparison
GPT-5.6 Terra vs Phi-4
GPT-5.6 Terra is the stronger model overall, scoring 59.2 to 31.2 on the Noometry Index. Phi-4 costs 51× less per token, which makes it the better buy when GPT-5.6 Terra's lead doesn't matter for your workload.
Last verified . 22 shared benchmarks.
MicrosoftSummary
- They share 22 benchmarks with published results for both. GPT-5.6 Terra scores higher in 9 categories and Phi-4 in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.6 Terra leads 81.6 to 20.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 99.7% for GPT-5.6 Terra and 13.8% for Phi-4.
- Phi-4 is cheaper at $0.07 / $0.14 per million input/output tokens, against $2 / $12 for GPT-5.6 Terra.
- GPT-5.6 Terra accepts more context: 1.05M tokens versus 128K.
- Phi-4 has downloadable open weights; the other is API-only.
Side by side
| GPT-5.6 Terra | Phi-4 | |
|---|---|---|
| Provider | OpenAI | Microsoft |
| Noometry Index | 59.2 | 31.2 |
| Released | 2026-07-09 | 2024-12-11 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 128K |
| Max output | 128K | 4K |
| Input $ / M tokens | $2 | $0.07 |
| Output $ / M tokens | $12 | $0.14 |
| Results tracked | 52 | 37 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.6 Terra leads
GPT-5.6 Terra: 57.7 (#19), Phi-4: 34.4 (#239)
| Benchmark | GPT-5.6 Terra | Phi-4 |
|---|---|---|
| LMArena Coding | 1484 | 1231 |
| DeepSWE | 69.6% | — |
| FrontierCode | 41.3% | — |
| CursorBench | 41.3% | — |
| LMArena WebDev | 1522 | — |
| SciCode | 55% | — |
| WeirdML | 78.3% | — |
| BigCodeBench Instruct | — | 45.5% |
| LiveBench Coding | — | 30.7% |
| BigCodeBench Complete | — | 55.4% |
| ALE-Bench | 1,951 | — |
Agentic & Tool Use GPT-5.6 Terra leads
GPT-5.6 Terra: 40.1 (#25), Phi-4: 22.8 (#128)
| Benchmark | GPT-5.6 Terra | Phi-4 |
|---|---|---|
| BALROG | 53.2% | 11.6% |
| APEX-Agents | 58.2% | — |
| Berkeley Function Calling Leaderboard | — | 28.8% |
| GDP.pdf | 24.7% | — |
| Vending-Bench 2 | 7,343 | — |
Reasoning GPT-5.6 Terra leads
GPT-5.6 Terra: 60.7 (#21), Phi-4: 17.7 (#291)
| Benchmark | GPT-5.6 Terra | Phi-4 |
|---|---|---|
| Chess Puzzles | 54% | 1% |
| LMArena Hard Prompts | 1468 | 1220 |
| Epoch Capabilities Index | 159.62 | 130.42 |
| ARC-AGI-2 | 83.9% | — |
| SimpleBench | 48.9% | — |
| Kagi LLM Benchmark | 51.3% | — |
| NYT Connections (extended) | 78.4% | — |
| ARC-AGI-1 | 96.5% | — |
| CritPt | 30% | — |
| LiveBench Reasoning | — | 47.8% |
| Mystery Game Puzzles | 35% | — |
| DTBench | 93.3% | — |
| LiveBench Data Analysis | — | 45.2% |
| LMCA | 55% | — |
| Surface Evolver Bench | 83.8% | — |
| LiveBench | — | 41.6% |
Math GPT-5.6 Terra leads
GPT-5.6 Terra: 81.6 (#12), Phi-4: 20.8 (#285)
| Benchmark | GPT-5.6 Terra | Phi-4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 99.7% | 13.8% |
| LMArena Math | 1466 | 1246 |
| FrontierMath (Tiers 1-3) | 86% | — |
| FrontierMath Tier 4 | 70.7% | — |
| ProofBench | 74% | — |
| LiveBench Math | — | 42% |
| MATH Level 5 | — | 64.9% |
Knowledge GPT-5.6 Terra leads
GPT-5.6 Terra: 61.2 (#30), Phi-4: 32.6 (#209)
| Benchmark | GPT-5.6 Terra | Phi-4 |
|---|---|---|
| GPQA Diamond | 93.3% | 56.1% |
| LMArena Expert | 1492 | 1203 |
| SimpleQA Verified | 43.2% | — |
| Confabulations | — | 29.4% |
| Vectara Hallucination Rate | — | 3.7% |
| MMLU | — | 84.8% |
Multimodal Not comparable
GPT-5.6 Terra: 47.3 (#11), Phi-4: —
| Benchmark | GPT-5.6 Terra | Phi-4 |
|---|---|---|
| LMArena Vision | 1271 | — |
| Blueprint-Bench 2 | 30.8% | — |
| Furniture Assembly | 54.2% | — |
| LMArena Document | 1472 | — |
Multilingual GPT-5.6 Terra leads
GPT-5.6 Terra: 54.4 (#44), Phi-4: 37.2 (#237)
| Benchmark | GPT-5.6 Terra | Phi-4 |
|---|---|---|
| LMArena Non-English | 1439 | 1197 |
| LMArena Chinese | 1513 | 1212 |
| LMArena French | 1471 | 1224 |
| LMArena German | 1460 | 1222 |
| LMArena Japanese | 1457 | 1158 |
| LMArena Korean | 1425 | 1151 |
| LMArena Russian | 1450 | 1209 |
| LMArena Spanish | 1448 | 1234 |
Instruction Following GPT-5.6 Terra leads
GPT-5.6 Terra: 76.4 (#40), Phi-4: 60.4 (#251)
| Benchmark | GPT-5.6 Terra | Phi-4 |
|---|---|---|
| LMArena Instruction Following | 1454 | 1201 |
| LiveBench Instruction Following | — | 58.4% |
Long Context GPT-5.6 Terra leads
GPT-5.6 Terra: 44.4 (#68), Phi-4: 36.9 (#226)
| Benchmark | GPT-5.6 Terra | Phi-4 |
|---|---|---|
| LMArena Longer Query | 1451 | 1217 |
Writing & Preference GPT-5.6 Terra leads
GPT-5.6 Terra: 70.2 (#23), Phi-4: 40.5 (#244)
| Benchmark | GPT-5.6 Terra | Phi-4 |
|---|---|---|
| LMArena Text | 1447 | 1217 |
| LMArena Creative Writing | 1410 | 1182 |
| LMArena Multi-Turn | 1449 | 1206 |
| Short-Story Creative Writing | — | 62.6% |
| EQ-Bench Creative Writing | 1855 | — |
| EQ-Bench 4 | 1234 | — |
| LiveBench Language | — | 25.6% |
Frequently asked questions
Is GPT-5.6 Terra better than Phi-4?
GPT-5.6 Terra is the stronger model overall, scoring 59.2 to 31.2 on the Noometry Index. Phi-4 costs 51× less per token, which makes it the better buy when GPT-5.6 Terra's lead doesn't matter for your workload.
Which is cheaper, GPT-5.6 Terra or Phi-4?
Phi-4 is cheaper. It lists at $0.07 per million input tokens and $0.14 per million output tokens; GPT-5.6 Terra lists at $2 and $12.
Is GPT-5.6 Terra or Phi-4 better for coding?
GPT-5.6 Terra scores higher on coding benchmarks: 57.7 versus 34.4 in the Noometry coding category.
Which has the bigger context window?
GPT-5.6 Terra does, with 1.05M tokens against 128K.
How many benchmarks do GPT-5.6 Terra and Phi-4 share?
22 benchmarks have published results for both models. GPT-5.6 Terra has 52 scored results on Noometry and Phi-4 has 37.