R1 0528 vs Hy3
Updated October 21, 2026. 39 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Pick R1 0528 for coding: it scores 71.4 on the Vibe Coding Index against 43.9 for Hy3.
Pick Hy3 when output tokens cost 4.1x less, or when the context window is larger, 262K against 164K tokens.
- Shared results
- 39
- R1 0528 only
- 20
- Hy3 only
- 5
- Like-for-like categories
- 6 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, four public coding benchmarks
R1 0528
R1 0528 scores 71.4 against 43.9.
One benchmark behind each score
Intelligence
AA Intelligence Index
No call
Only R1 0528 has a AA Intelligence Index result, 13.1.
Directional only
Web design
Design Arena: websites
Hy3
Hy3 leads, 1230 against 1149.
Same board, both measured
Long documents
Listed context window
Hy3
Hy3 takes 262K tokens against 164K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Hy3
R1 0528 $0.0016, Hy3 $0.0004 per turn.
List prices
At a glance
- Coding index
- R1 0528 71.4 from 1 benchmark, Hy3 43.9 from 1 benchmark.
- Widest category
- Web design, 12.2 points apart on average over 7 shared benchmarks.
- Widest single gap
- Long context reasoning (AA-LCR): R1 0528 55.7% against Hy3 79.0%, 23.3 points apart on a 0 to 100 scale.
- Output price
- R1 0528 $2.15 against Hy3 $0.528 per million output tokens, 4.1x apart.
- Context window
- R1 0528 takes 164K 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 | R1 0528 | Bars | Hy3 | Ties |
|---|---|---|---|---|
| Coding | 0 | 3 | ||
| Reasoning | 0 | 5 | ||
| Math | 0 | 2 | ||
| Knowledge | 0 | 4 | ||
| Web design | 1 | 6 | ||
| General | 2 | 16 | ||
| All shared boards | 3 | 36 |
CodingLike-for-likeFour of these boards feed the Vibe Coding Index03
| Benchmark | R1 0528 | Hy3 | Leader |
|---|---|---|---|
| Aider polyglot | 71.4% | Not measured | Only R1 0528 measured |
| LMArena WebDev | Not measured | 1507 | Only Hy3 measured |
| LiveCodeBench (Artificial Analysis run) | 77.0% | Not measured | Only R1 0528 measured |
| SciCode | Not measured | 48.6% | Only Hy3 measured |
| Terminal-Bench Hard (Artificial Analysis run) | 15.9% | Not measured | Only R1 0528 measured |
| WeirdML | 41.6% | Not measured | Only R1 0528 measured |
| ALE-Bench | 804.1 | 838.6 | Hy3 |
| LMArena Text: coding | 1427 | 1466 | Hy3 |
| LMArena Text: software and IT | 1435 | 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 27 boards here | |||
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index1 board, none shared
| Benchmark | R1 0528 | Hy3 | Leader |
|---|---|---|---|
| AA Intelligence Index | 13.1 | Not measured | Only R1 0528 measured |
| Neither model has a result on the other 1 board here | |||
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index3 boards, none shared
| Benchmark | R1 0528 | Hy3 | Leader |
|---|---|---|---|
| τ²-Bench Telecom | 36.5% | Not measured | Only R1 0528 measured |
| METR time horizons | 53.8% | Not measured | Only R1 0528 measured |
| DeepResearch Bench | 35.1% | Not measured | Only R1 0528 measured |
| Neither model has a result on the other 36 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index05
| Benchmark | R1 0528 | Hy3 | Leader |
|---|---|---|---|
| GPQA Diamond (Artificial Analysis run) | 81.3% | 89.7% | Hy3 |
| Humanity's Last Exam (Artificial Analysis run) | 15.8% | 33.5% | Hy3 |
| GPQA Diamond (Epoch AI run) | 76.3% | Not measured | Only R1 0528 measured |
| ARC-AGI | 21.2% | Not measured | Only R1 0528 measured |
| ARC-AGI-2 | 1.1% | Not measured | Only R1 0528 measured |
| SimpleBench | 40.8% | Not measured | Only R1 0528 measured |
| LMArena Text: hard prompts | 1415 | 1447 | Hy3 |
| LMArena Text: expert | 1394 | 1461 | Hy3 |
| LMArena Text: hard prompts in English | 1422 | 1455 | Hy3 |
| Neither model has a result on the other 12 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | R1 0528 | Hy3 | Leader |
|---|---|---|---|
| AA Math Index | 76.0 | Not measured | Only R1 0528 measured |
| AIME 2025 (Artificial Analysis run) | 76.0% | Not measured | Only R1 0528 measured |
| OTIS Mock AIME | 66.4% | Not measured | Only R1 0528 measured |
| MATH level 5 | 96.6% | Not measured | Only R1 0528 measured |
| LMArena Text: math | 1400 | 1473 | Hy3 |
| LMArena Text: mathematical work | 1415 | 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 | R1 0528 | Hy3 | Leader |
|---|---|---|---|
| MMLU Pro (Artificial Analysis run) | 84.9% | Not measured | Only R1 0528 measured |
| LMArena Text: science | 1430 | 1457 | Hy3 |
| LMArena Text: medicine and healthcare | 1445 | 1452 | Hy3 |
| LMArena Text: legal and government | 1419 | 1422 | Hy3 |
| LMArena Text: business and finance | 1402 | 1424 | Hy3 |
| Neither model has a result on the other 12 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index16
| Benchmark | R1 0528 | Hy3 | Leader |
|---|---|---|---|
| Design Arena: websites | 1149 | 1230 | Hy3 |
| Design Arena: UI components | 1102 | 1205 | Hy3 |
| Design Arena: game dev | 1098 | 1208 | Hy3 |
| Design Arena: data viz | 1182 | 1163 | R1 0528 |
| Design Arena: 3D | 1111 | 1233 | Hy3 |
| Design Arena: all code categories | 1139 | 1227 | Hy3 |
| Design Arena: SVG | 1038 | 1199 | Hy3 |
| 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 index216
| Benchmark | R1 0528 | Hy3 | Leader |
|---|---|---|---|
| IFBench | 39.6% | Not measured | Only R1 0528 measured |
| Long context reasoning (AA-LCR) | 55.7% | 79.0% | Hy3 |
| Fiction.LiveBench | 75.0% | Not measured | Only R1 0528 measured |
| Creative writing (Lech Mazur) | 81.9% | Not measured | Only R1 0528 measured |
| LMArena Text | 1428 | 1438 | Hy3 |
| LMArena Text: instruction following | 1382 | 1426 | Hy3 |
| LMArena Text: creative writing | 1405 | 1403 | R1 0528 |
| LMArena Text: English | 1437 | 1448 | Hy3 |
| LMArena Text: non-English | 1412 | 1426 | Hy3 |
| LMArena Text: Chinese | 1442 | 1485 | Hy3 |
| LMArena Text: German | 1404 | 1449 | Hy3 |
| LMArena Text: Japanese | 1391 | 1397 | Hy3 |
| LMArena Text: Korean | 1360 | 1396 | Hy3 |
| LMArena Text: French | 1417 | 1456 | Hy3 |
| LMArena Text: Spanish | 1411 | 1457 | Hy3 |
| LMArena Text: Russian | 1423 | 1432 | Hy3 |
| LMArena Text: Polish | 1428 | 1447 | Hy3 |
| LMArena Text: multi-turn | 1405 | 1436 | Hy3 |
| LMArena Text: longer queries | 1391 | 1442 | Hy3 |
| LMArena Text: writing and language | 1402 | 1414 | Hy3 |
| LMArena Text: entertainment and media | 1400 | 1400 | R1 0528 |
| Neither model has a result on the other 25 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | R1 0528 | 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
Hy3 costs less
Repository review
50K input + 3K output tokens
Hy3 costs less
Agent loop
200K input + 10K output tokens
Hy3 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 | DeepSeek R1 0528 | Tencent Hy3 |
|---|---|---|
| Maker | DeepSeek | Tencent |
| Weights | Open weights | Open weights |
| Released | May 28, 2025 | July 6, 2026 |
| Context window | 164K | 262K |
| Input $/Mtok | $0.50 | $0.132 |
| Output $/Mtok | $2.15 | $0.528 |
| Coding index | 71.4 | 43.9 |
| Independent measurements behind it | 1 | 1 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is R1 0528 better than Hy3 for coding?
Pick R1 0528 for coding: it scores 71.4 on the Vibe Coding Index against 43.9 for Hy3. Pick Hy3 when output tokens cost 4.1x less, or when the context window is larger, 262K against 164K tokens.
Which is cheaper, R1 0528 or Hy3?
R1 0528 costs $0.50 per million input tokens and $2.15 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, R1 0528 or Hy3?
Hy3, with 262K tokens against 164K.
How many benchmarks is the R1 0528 vs Hy3 comparison based on?
39 benchmarks have a result for both models as of October 21, 2026. R1 0528's coding index rests on 1 independent benchmark and Hy3's on 1. Where only one model has a result, the page says so instead of naming a winner.