Phi 4 vs Hy3
Updated October 21, 2026. 30 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Neither model has a coding score yet, so the two cannot be ranked on coding.
Phi 4 has the edge where output tokens cost 3.8x less, and where what matters is LMArena Text: coding or LMArena Text: software and IT, where it leads.
- Shared results
- 30
- Phi 4 only
- 15
- Hy3 only
- 14
- Like-for-like categories
- 5 of 9
Shared results can support a head-to-head reading. Results only one model has describe coverage, not superiority.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
No call
Neither has a coding score yet.
One side not measured
Intelligence
AA Intelligence Index
No call
Only Phi 4 has a AA Intelligence Index result, 5.9.
Directional only
Web design
Design Arena: websites
No call
Only Hy3 has a Design Arena: websites result, 1230.
Directional only
Long documents
Listed context window
Hy3
Hy3 takes 262K tokens against 16K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Phi 4
Phi 4 $0.0001, Hy3 $0.0004 per turn.
List prices
At a glance
- Coding index
- Phi 4 not benchmarked, Hy3 not benchmarked.
- Widest category
- General, 40.8 points apart on average over 17 shared benchmarks.
- Widest single gap
- Long context reasoning (AA-LCR): Phi 4 0.0% against Hy3 79.0%, 79.0 points apart on a 0 to 100 scale.
- Output price
- Phi 4 $0.14 against Hy3 $0.528 per million output tokens, 3.8x apart.
- Context window
- Phi 4 takes 16K tokens, Hy3 262K.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | Phi 4 | Bars | Hy3 | Ties |
|---|---|---|---|---|
| Coding | 0 | 2 | ||
| Reasoning | 0 | 5 | ||
| Math | 0 | 2 | ||
| Knowledge | 0 | 4 | ||
| General | 0 | 17 | ||
| All shared boards | 0 | 30 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index02
| Benchmark | Phi 4 | Hy3 | Leader |
|---|---|---|---|
| LMArena WebDev | Not measured | 1507 | Only Hy3 measured |
| LiveCodeBench (Artificial Analysis run) | 23.1% | Not measured | Only Phi 4 measured |
| SciCode | Not measured | 48.6% | Only Hy3 measured |
| Terminal-Bench Hard (Artificial Analysis run) | 3.8% | Not measured | Only Phi 4 measured |
| ALE-Bench | Not measured | 838.6 | Only Hy3 measured |
| LMArena Text: coding | 1231 | 1466 | Hy3 |
| LMArena Text: software and IT | 1228 | 1463 | Hy3 |
| LMArena WebDev: React | Not measured | 1511 | Only Hy3 measured |
| LMArena WebDev: HTML | Not measured | 1487 | Only Hy3 measured |
| Neither model has a result on the other 29 boards here | |||
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
| Benchmark | Phi 4 | Hy3 | Leader |
|---|---|---|---|
| AA Intelligence Index | 5.9 | Not measured | Only Phi 4 measured |
| Epoch Capabilities Index | 130.5 | Not measured | Only Phi 4 measured |
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
| Benchmark | Phi 4 | Hy3 | Leader |
|---|---|---|---|
| τ²-Bench Telecom | 0.0% | Not measured | Only Phi 4 measured |
| BALROG | 11.6% | Not measured | Only Phi 4 measured |
| Neither model has a result on the other 37 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index05
| Benchmark | Phi 4 | Hy3 | Leader |
|---|---|---|---|
| GPQA Diamond (Artificial Analysis run) | 57.5% | 89.7% | Hy3 |
| Humanity's Last Exam (Artificial Analysis run) | 3.8% | 33.5% | Hy3 |
| GPQA Diamond (Epoch AI run) | 56.1% | Not measured | Only Phi 4 measured |
| LMArena Text: hard prompts | 1220 | 1447 | Hy3 |
| LMArena Text: expert | 1203 | 1461 | Hy3 |
| LMArena Text: hard prompts in English | 1230 | 1455 | Hy3 |
| Neither model has a result on the other 15 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Phi 4 | Hy3 | Leader |
|---|---|---|---|
| AA Math Index | 18.0 | Not measured | Only Phi 4 measured |
| AIME 2025 (Artificial Analysis run) | 18.0% | Not measured | Only Phi 4 measured |
| OTIS Mock AIME | 13.8% | Not measured | Only Phi 4 measured |
| MATH level 5 | 64.9% | Not measured | Only Phi 4 measured |
| LMArena Text: math | 1246 | 1473 | Hy3 |
| LMArena Text: mathematical work | 1254 | 1482 | Hy3 |
| Neither model has a result on the other 7 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index04
| Benchmark | Phi 4 | Hy3 | Leader |
|---|---|---|---|
| MMLU Pro (Artificial Analysis run) | 71.4% | Not measured | Only Phi 4 measured |
| MMLU | 84.8% | Not measured | Only Phi 4 measured |
| LMArena Text: science | 1206 | 1457 | Hy3 |
| LMArena Text: medicine and healthcare | 1202 | 1452 | Hy3 |
| LMArena Text: legal and government | 1241 | 1422 | Hy3 |
| LMArena Text: business and finance | 1199 | 1424 | Hy3 |
| Neither model has a result on the other 11 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
| Benchmark | Phi 4 | Hy3 | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1230 | Only Hy3 measured |
| Design Arena: UI components | Not measured | 1207 | Only Hy3 measured |
| Design Arena: game dev | Not measured | 1208 | Only Hy3 measured |
| Design Arena: data viz | Not measured | 1163 | Only Hy3 measured |
| Design Arena: 3D | Not measured | 1233 | Only Hy3 measured |
| Design Arena: all code categories | Not measured | 1227 | Only Hy3 measured |
| Design Arena: SVG | Not measured | 1200 | Only Hy3 measured |
| Design Arena: ASCII art | Not measured | 1157 | Only Hy3 measured |
| Neither model has a result on the other 1 board here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index017
| Benchmark | Phi 4 | Hy3 | Leader |
|---|---|---|---|
| IFBench | 23.5% | Not measured | Only Phi 4 measured |
| Long context reasoning (AA-LCR) | 0.0% | 79.0% | Hy3 |
| Creative writing (Lech Mazur) | 62.6% | Not measured | Only Phi 4 measured |
| LMArena Text | 1217 | 1438 | Hy3 |
| LMArena Text: instruction following | 1201 | 1426 | Hy3 |
| LMArena Text: creative writing | 1181 | 1403 | Hy3 |
| LMArena Text: English | 1231 | 1448 | Hy3 |
| LMArena Text: non-English | 1197 | 1426 | Hy3 |
| LMArena Text: Chinese | 1212 | 1485 | Hy3 |
| LMArena Text: German | 1222 | 1449 | Hy3 |
| LMArena Text: Japanese | 1158 | 1397 | Hy3 |
| LMArena Text: Korean | 1151 | 1396 | Hy3 |
| LMArena Text: French | 1224 | 1456 | Hy3 |
| LMArena Text: Spanish | 1234 | 1457 | Hy3 |
| LMArena Text: Russian | 1209 | 1432 | Hy3 |
| LMArena Text: Polish | Not measured | 1447 | Only Hy3 measured |
| LMArena Text: multi-turn | 1206 | 1436 | Hy3 |
| LMArena Text: longer queries | 1217 | 1442 | Hy3 |
| LMArena Text: writing and language | 1198 | 1414 | Hy3 |
| LMArena Text: entertainment and media | 1162 | 1400 | Hy3 |
| Neither model has a result on the other 26 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Phi 4 | Hy3 | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
Phi 4 costs less
Repository review
50K input + 3K output tokens
Phi 4 costs less
Agent loop
200K input + 10K output tokens
Phi 4 costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | Microsoft Phi 4 | Tencent Hy3 |
|---|---|---|
| Maker | Microsoft | Tencent |
| Weights | Open weights | Open weights |
| Released | January 10, 2025 | July 6, 2026 |
| Context window | 16K | 262K |
| Input $/Mtok | $0.07 | $0.132 |
| Output $/Mtok | $0.14 | $0.528 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 0 | 1 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Phi 4 better than Hy3 for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. Phi 4 has the edge where output tokens cost 3.8x less, and where what matters is LMArena Text: coding or LMArena Text: software and IT, where it leads.
Which is cheaper, Phi 4 or Hy3?
Phi 4 costs $0.07 per million input tokens and $0.14 per million output tokens; Hy3 costs $0.132 and $0.528. Cost per solved task is known for neither of them.
Which has the bigger context window, Phi 4 or Hy3?
Hy3, with 262K tokens against 16K.
How many benchmarks is the Phi 4 vs Hy3 comparison based on?
30 benchmarks have a result for both models as of October 21, 2026. Phi 4's coding index rests on 0 independent benchmarks and Hy3's on 1. Where only one model has a result, the page says so instead of naming a winner.