# Claude Sonnet 4.6 vs Phi-4

> Claude Sonnet 4.6 is the stronger model overall, scoring 50.3 to 31.2 on the Noometry Index. Phi-4 costs 69× less per token, which makes it the better buy when Claude Sonnet 4.6's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/claude-sonnet-4-6-vs-phi-4
- Last updated: 2026-10-11
- Shared benchmarks: 22

## Summary

- They share 22 benchmarks with published results for both. Claude Sonnet 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 math, where Claude Sonnet 4.6 leads 52.9 to 20.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 85.8% for Claude Sonnet 4.6 and 13.8% for Phi-4.
- Phi-4 is cheaper at $0.07 / $0.14 per million input/output tokens, against $3 / $15 for Claude Sonnet 4.6.
- Claude Sonnet 4.6 accepts more context: 1M tokens versus 128K.
- Phi-4 has downloadable open weights; the other is API-only.

## Snapshot

| | Claude Sonnet 4.6 | Phi-4 |
|---|---|---|
| Provider | Anthropic | Microsoft |
| Noometry Index | 50.3 | 31.2 |
| Rank | 50 | 279 |
| Context | 1M | 128K |
| Input $/M | $3 | $0.07 |
| Output $/M | $15 | $0.14 |
| Weights | Proprietary | Open |

## Coding

- Claude Sonnet 4.6: 46.3 (#67)
- Phi-4: 34.4 (#239)

| Benchmark | Claude Sonnet 4.6 | Phi-4 |
|---|---|---|
| LMArena Coding | 1504 | 1231 |
| SWE-bench Verified | 75.2% | — |
| DeepSWE | 29.9% | — |
| FrontierCode | 24.3% | — |
| LMArena WebDev | 1522 | — |
| SciCode | 46.8% | — |
| WeirdML | 66.1% | — |
| BigCodeBench Instruct | — | 45.5% |
| LiveBench Coding | — | 30.7% |
| BigCodeBench Complete | — | 55.4% |
| ALE-Bench | 1,327 | — |

## Agentic & Tool Use

- Claude Sonnet 4.6: 39.1 (#28)
- Phi-4: 22.8 (#128)

| Benchmark | Claude Sonnet 4.6 | Phi-4 |
|---|---|---|
| Terminal-Bench | 53.4% | — |
| APEX-Agents | 43% | — |
| Berkeley Function Calling Leaderboard | — | 28.8% |
| OSWorld 2.0 | 9.3% | — |
| DeepResearch Bench | 54.9% | — |
| OSWorld | 72.1% | — |
| BALROG | — | 11.6% |
| ExploitBench | 23.6% | — |
| GBAEval | 48.8% | — |
| GDP.pdf | 18% | — |
| LMArena Search | 1221 | — |
| Vending-Bench 2 | 7,204 | — |

## Reasoning

- Claude Sonnet 4.6: 46.1 (#45)
- Phi-4: 17.7 (#291)

| Benchmark | Claude Sonnet 4.6 | Phi-4 |
|---|---|---|
| Chess Puzzles | 13% | 1% |
| LMArena Hard Prompts | 1484 | 1220 |
| Epoch Capabilities Index | 152.24 | 130.42 |
| ARC-AGI-2 | 60.4% | — |
| NYT Connections (extended) | 80.9% | — |
| ARC-AGI-1 | 86.5% | — |
| CritPt | 3.1% | — |
| Thematic Generalization | 76.3% | — |
| LiveBench Reasoning | — | 47.8% |
| Mystery Game Puzzles | 16% | — |
| DTBench | 89.9% | — |
| LiveBench Data Analysis | — | 45.2% |
| LMCA | 46.5% | — |
| ForecastBench | 62 | — |
| LiveBench | — | 41.6% |

## Math

- Claude Sonnet 4.6: 52.9 (#49)
- Phi-4: 20.8 (#285)

| Benchmark | Claude Sonnet 4.6 | Phi-4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 85.8% | 13.8% |
| LMArena Math | 1462 | 1246 |
| ProofBench | 45% | — |
| LiveBench Math | — | 42% |
| MATH Level 5 | — | 64.9% |
| FrontierMath (Feb 2025 set) | 32.4% | — |
| FrontierMath Tier 4 (v1) | 8.3% | — |

## Knowledge

- Claude Sonnet 4.6: 51.7 (#65)
- Phi-4: 32.6 (#209)

| Benchmark | Claude Sonnet 4.6 | Phi-4 |
|---|---|---|
| GPQA Diamond | 87.4% | 56.1% |
| Vectara Hallucination Rate | 10.6% | 3.7% |
| LMArena Expert | 1500 | 1203 |
| SimpleQA Verified | 35.5% | — |
| Confabulations | — | 29.4% |
| MMLU | — | 84.8% |

## Multimodal

- Claude Sonnet 4.6: 38.0 (#68)
- Phi-4: —

| Benchmark | Claude Sonnet 4.6 | Phi-4 |
|---|---|---|
| LMArena Vision | 1283 | — |
| Blueprint-Bench 2 | 6.7% | — |
| LMArena Document | 1482 | — |

## Multilingual

- Claude Sonnet 4.6: 54.4 (#41)
- Phi-4: 37.2 (#237)

| Benchmark | Claude Sonnet 4.6 | Phi-4 |
|---|---|---|
| LMArena Non-English | 1440 | 1197 |
| LMArena Chinese | 1491 | 1212 |
| LMArena French | 1465 | 1224 |
| LMArena German | 1428 | 1222 |
| LMArena Japanese | 1420 | 1158 |
| LMArena Korean | 1411 | 1151 |
| LMArena Russian | 1440 | 1209 |
| LMArena Spanish | 1464 | 1234 |

## Instruction Following

- Claude Sonnet 4.6: 77.4 (#25)
- Phi-4: 60.4 (#251)

| Benchmark | Claude Sonnet 4.6 | Phi-4 |
|---|---|---|
| LMArena Instruction Following | 1475 | 1201 |
| LiveBench Instruction Following | — | 58.4% |

## Long Context

- Claude Sonnet 4.6: 45.3 (#44)
- Phi-4: 36.9 (#226)

| Benchmark | Claude Sonnet 4.6 | Phi-4 |
|---|---|---|
| LMArena Longer Query | 1479 | 1217 |

## Writing & Preference

- Claude Sonnet 4.6: 70.2 (#22)
- Phi-4: 40.5 (#244)

| Benchmark | Claude Sonnet 4.6 | Phi-4 |
|---|---|---|
| LMArena Text | 1458 | 1217 |
| LMArena Creative Writing | 1435 | 1182 |
| LMArena Multi-Turn | 1464 | 1206 |
| Short-Story Creative Writing | — | 62.6% |
| EQ-Bench Creative Writing | 1810 | — |
| EQ-Bench 4 | 1207 | — |
| LiveBench Language | — | 25.6% |

## FAQ

### Is Claude Sonnet 4.6 better than Phi-4?

Claude Sonnet 4.6 is the stronger model overall, scoring 50.3 to 31.2 on the Noometry Index. Phi-4 costs 69× less per token, which makes it the better buy when Claude Sonnet 4.6's lead doesn't matter for your workload.

### Which is cheaper, Claude Sonnet 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; Claude Sonnet 4.6 lists at $3 and $15.

### Is Claude Sonnet 4.6 or Phi-4 better for coding?

Claude Sonnet 4.6 scores higher on coding benchmarks: 46.3 versus 34.4 in the Noometry coding category.

### Which has the bigger context window?

Claude Sonnet 4.6 does, with 1M tokens against 128K.

### How many benchmarks do Claude Sonnet 4.6 and Phi-4 share?

22 benchmarks have published results for both models. Claude Sonnet 4.6 has 57 scored results on Noometry and Phi-4 has 37.