Model comparison
GPT-6 Sol vs Phi-4
GPT-6 Sol is the stronger model overall, scoring 61.8 to 31.2 on the Noometry Index. Phi-4 costs 46× less per token, which makes it the better buy when GPT-6 Sol's lead doesn't matter for your workload.
Last verified . 21 shared benchmarks.
MicrosoftSummary
- They share 21 benchmarks with published results for both. GPT-6 Sol 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-6 Sol leads 87.2 to 20.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 100% for GPT-6 Sol and 13.8% for Phi-4.
- Phi-4 is cheaper at $0.07 / $0.14 per million input/output tokens, against $2 / $10 for GPT-6 Sol.
- GPT-6 Sol accepts more context: 1.05M tokens versus 128K.
- Phi-4 has downloadable open weights; the other is API-only.
Side by side
| GPT-6 Sol | Phi-4 | |
|---|---|---|
| Provider | OpenAI | Microsoft |
| Noometry Index | 61.8 | 31.2 |
| Released | 2026-09-22 | 2024-12-11 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 128K |
| Max output | 128K | 4K |
| Input $ / M tokens | $2 | $0.07 |
| Output $ / M tokens | $10 | $0.14 |
| Results tracked | 45 | 37 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-6 Sol leads
GPT-6 Sol: 60.1 (#11), Phi-4: 34.4 (#239)
| Benchmark | GPT-6 Sol | Phi-4 |
|---|---|---|
| LMArena Coding | 1447 | 1231 |
| DeepSWE | 68.8% | — |
| FrontierCode | 49.3% | — |
| LMArena WebDev | 1688 | — |
| SciCode | 57.6% | — |
| BigCodeBench Instruct | — | 45.5% |
| LiveBench Coding | — | 30.7% |
| BigCodeBench Complete | — | 55.4% |
| ALE-Bench | 2,462 | — |
Agentic & Tool Use GPT-6 Sol leads
GPT-6 Sol: 37.2 (#36), Phi-4: 22.8 (#128)
| Benchmark | GPT-6 Sol | Phi-4 |
|---|---|---|
| APEX-Agents | 54.3% | — |
| Berkeley Function Calling Leaderboard | — | 28.8% |
| BALROG | — | 11.6% |
| GDP.pdf | 26.4% | — |
| Vending-Bench 2 | 14,428 | — |
Reasoning GPT-6 Sol leads
GPT-6 Sol: 74.0 (#9), Phi-4: 17.7 (#291)
| Benchmark | GPT-6 Sol | Phi-4 |
|---|---|---|
| LMArena Hard Prompts | 1418 | 1220 |
| Epoch Capabilities Index | 162.72 | 130.42 |
| ARC-AGI-2 | 89.6% | — |
| NYT Connections (extended) | 90.1% | — |
| ARC-AGI-1 | 95.5% | — |
| CritPt | 30.9% | — |
| Chess Puzzles | — | 1% |
| EBR-Bench | 53.3% | — |
| LiveBench Reasoning | — | 47.8% |
| Mystery Game Puzzles | 56% | — |
| DTBench | 97.3% | — |
| LiveBench Data Analysis | — | 45.2% |
| LMCA | 59.1% | — |
| LiveBench | — | 41.6% |
Math GPT-6 Sol leads
GPT-6 Sol: 87.2 (#7), Phi-4: 20.8 (#285)
| Benchmark | GPT-6 Sol | Phi-4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 100% | 13.8% |
| LMArena Math | 1402 | 1246 |
| FrontierMath (Tiers 1-3) | 89.8% | — |
| FrontierMath Tier 4 | 90% | — |
| ProofBench | 83% | — |
| LiveBench Math | — | 42% |
| MATH Level 5 | — | 64.9% |
Knowledge GPT-6 Sol leads
GPT-6 Sol: 64.8 (#15), Phi-4: 32.6 (#209)
| Benchmark | GPT-6 Sol | Phi-4 |
|---|---|---|
| GPQA Diamond | 94.3% | 56.1% |
| Vectara Hallucination Rate | 6.5% | 3.7% |
| LMArena Expert | 1439 | 1203 |
| SimpleQA Verified | 60.7% | — |
| Confabulations | — | 29.4% |
| MMLU | — | 84.8% |
Multimodal Not comparable
GPT-6 Sol: 47.6 (#10), Phi-4: —
| Benchmark | GPT-6 Sol | Phi-4 |
|---|---|---|
| LMArena Vision | 1245 | — |
| Blueprint-Bench 2 | 36.9% | — |
| Furniture Assembly | 58.3% | — |
Multilingual GPT-6 Sol leads
GPT-6 Sol: 50.5 (#118), Phi-4: 37.2 (#237)
| Benchmark | GPT-6 Sol | Phi-4 |
|---|---|---|
| LMArena Non-English | 1385 | 1197 |
| LMArena Chinese | 1405 | 1212 |
| LMArena French | 1410 | 1224 |
| LMArena German | 1390 | 1222 |
| LMArena Japanese | 1385 | 1158 |
| LMArena Korean | 1341 | 1151 |
| LMArena Russian | 1401 | 1209 |
| LMArena Spanish | 1384 | 1234 |
Instruction Following GPT-6 Sol leads
GPT-6 Sol: 74.5 (#94), Phi-4: 60.4 (#251)
| Benchmark | GPT-6 Sol | Phi-4 |
|---|---|---|
| LMArena Instruction Following | 1412 | 1201 |
| LiveBench Instruction Following | — | 58.4% |
Long Context GPT-6 Sol leads
GPT-6 Sol: 43.1 (#108), Phi-4: 36.9 (#226)
| Benchmark | GPT-6 Sol | Phi-4 |
|---|---|---|
| LMArena Longer Query | 1411 | 1217 |
Writing & Preference GPT-6 Sol leads
GPT-6 Sol: 71.9 (#18), Phi-4: 40.5 (#244)
| Benchmark | GPT-6 Sol | Phi-4 |
|---|---|---|
| LMArena Text | 1395 | 1217 |
| LMArena Creative Writing | 1378 | 1182 |
| LMArena Multi-Turn | 1412 | 1206 |
| Short-Story Creative Writing | — | 62.6% |
| EQ-Bench Creative Writing | 2125 | — |
| LiveBench Language | — | 25.6% |
Frequently asked questions
Is GPT-6 Sol better than Phi-4?
GPT-6 Sol is the stronger model overall, scoring 61.8 to 31.2 on the Noometry Index. Phi-4 costs 46× less per token, which makes it the better buy when GPT-6 Sol's lead doesn't matter for your workload.
Which is cheaper, GPT-6 Sol or Phi-4?
Phi-4 is cheaper. It lists at $0.07 per million input tokens and $0.14 per million output tokens; GPT-6 Sol lists at $2 and $10.
Is GPT-6 Sol or Phi-4 better for coding?
GPT-6 Sol scores higher on coding benchmarks: 60.1 versus 34.4 in the Noometry coding category.
Which has the bigger context window?
GPT-6 Sol does, with 1.05M tokens against 128K.
How many benchmarks do GPT-6 Sol and Phi-4 share?
21 benchmarks have published results for both models. GPT-6 Sol has 45 scored results on Noometry and Phi-4 has 37.