Grok 4.3 vs Hy3 preview
Hy3 preview leads Grok 4.3 on the Artificial Analysis Intelligence Index, 25.3 to 24.9, and Hy3 preview costs 4.2x less per output token ($0.60 vs $2.50 per million). Updated October 21, 2026; 36 benchmarks have a result for both.
Grok 4.3 or Hy3 preview: which to pick
Only Grok 4.3 has a coding score, 37.4 on the Vibe Coding Index; Hy3 preview has none, so the two cannot be ranked on coding.
Pick Hy3 preview for output tokens 4.2x cheaper ($0.60 vs $2.50 per million) and work measured by AA Coding Index or LMArena Text: coding, where it leads.
- Shared results
- 36
- Grok 4.3 only
- 68
- Hy3 preview only
- 0
- Like-for-like categories
- 7 of 9
Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.
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
Only Grok 4.3 has a coding score, so no call is made.
One side not measured
Intelligence
AA Intelligence Index
Hy3 preview
Hy3 preview leads, 25.3 against 24.9.
Same board, both measured
Agentic work
AA Agentic Index
Hy3 preview
Hy3 preview leads, 24.1 against 15.5.
Same board, both measured
Reasoning
GPQA Diamond
No call
Only Grok 4.3 has a GPQA Diamond result, 91.4%.
Directional only
Knowledge
MMLU Pro
No call
Only Grok 4.3 has a MMLU Pro result, 85.8%.
Directional only
Web design
Design Arena: websites
No call
Only Grok 4.3 has a Design Arena: websites result, 1196.
Directional only
Long documents
Listed context window
Grok 4.3
Grok 4.3 takes 1M tokens against 262K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Hy3 preview
Grok 4.3 $0.0025, Hy3 preview $0.0005 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only Grok 4.3 has the number, $0.31.
Directional only
At a glance
- Coding index
- Grok 4.3 37.4 from 3 benchmarks, Hy3 preview not benchmarked.
- Widest category
- Math, 6.9 points apart on average over 2 shared benchmarks.
- Widest single gap
- IFBench: Grok 4.3 83.3% against Hy3 preview 63.1%, 20.2 points apart on a 0 to 100 scale.
- Output price
- Grok 4.3 $2.50 against Hy3 preview $0.60 per million output tokens, 4.2x apart.
- Context window
- Grok 4.3 takes 1M tokens, Hy3 preview 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 | Grok 4.3 | Bars | Hy3 preview | Ties |
|---|---|---|---|---|
| Coding | 4 | 3 | ||
| Intelligence | 0 | 1 | ||
| Agentic work | 1 | 1 | ||
| Reasoning | 2 | 3 | ||
| Math | 0 | 2 | ||
| Knowledge | 0 | 4 | ||
| General | 6 | 9 | ||
| All shared boards | 13 | 23 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index43
| Benchmark | Grok 4.3 | Hy3 preview | Leader |
|---|---|---|---|
| LMArena WebDev | 1358 | 1357 | Grok 4.3 |
| AA Coding Index | 42.2 | 58.8 | Hy3 preview |
| LiveCodeBench (Vals AI) | 84.5% | Not measured | Only Grok 4.3 measured |
| SWE-bench Verified (Vals AI) | 71.4% | Not measured | Only Grok 4.3 measured |
| Vibe Code Bench (Vals AI) | 19.4% | Not measured | Only Grok 4.3 measured |
| Code migration (Vals AI) | 6.8% | Not measured | Only Grok 4.3 measured |
| ProgramBench (Vals AI) | 0.0% | Not measured | Only Grok 4.3 measured |
| Terminal-Bench 2.1 (Vals AI) | 42.0% | Not measured | Only Grok 4.3 measured |
| SkillsBench (Vals AI) | 40.6% | Not measured | Only Grok 4.3 measured |
| SciCode | 48.3% | Not measured | Only Grok 4.3 measured |
| Terminal-Bench Hard (Artificial Analysis run) | 37.9% | 34.1% | Grok 4.3 |
| WeirdML | 49.9% | Not measured | Only Grok 4.3 measured |
| ALE-Bench | 944.2 | Not measured | Only Grok 4.3 measured |
| LMArena Text: coding | 1415 | 1421 | Hy3 preview |
| LMArena Text: software and IT | 1416 | 1423 | Hy3 preview |
| LMArena WebDev: React | 1353 | 1349 | Grok 4.3 |
| LMArena WebDev: HTML | 1367 | 1356 | Grok 4.3 |
| LMArena WebDev: from an image | 1375 | Not measured | Only Grok 4.3 measured |
| Neither model has a result on the other 21 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Grok 4.3 | Hy3 preview | Leader |
|---|---|---|---|
| AA Intelligence Index | 24.9 | 25.3 | Hy3 preview |
| Neither model has a result on the other 1 board here | |||
Agentic workLike-for-likePublic boards, shown as published, kept out of the index11
| Benchmark | Grok 4.3 | Hy3 preview | Leader |
|---|---|---|---|
| AA Agentic Index | 15.5 | 24.1 | Hy3 preview |
| Design Arena: agents building web apps | 1126 | Not measured | Only Grok 4.3 measured |
| Design Arena: agents building full-stack apps | 1006 | Not measured | Only Grok 4.3 measured |
| Design Arena: agents building mobile apps | 1059 | Not measured | Only Grok 4.3 measured |
| Design Arena: agents building Android apps | 961 | Not measured | Only Grok 4.3 measured |
| Design Arena: agents building Godot games | 914 | Not measured | Only Grok 4.3 measured |
| Design Arena: agents building games | 1007 | Not measured | Only Grok 4.3 measured |
| Design Arena: HTML slides | 1008 | Not measured | Only Grok 4.3 measured |
| Design Arena: PowerPoint slides | 1013 | Not measured | Only Grok 4.3 measured |
| Finance Agent (Vals AI) | 37.7% | Not measured | Only Grok 4.3 measured |
| Tax Agent Bench (Vals AI) | 6.2% | Not measured | Only Grok 4.3 measured |
| Legal Research Bench (Vals AI) | 15.4% | Not measured | Only Grok 4.3 measured |
| Excel Modeling Benchmark (Vals AI) | 18.4% | Not measured | Only Grok 4.3 measured |
| τ²-Bench Telecom | 97.7% | 92.7% | Grok 4.3 |
| Vending-Bench 2 | $35 | Not measured | Only Grok 4.3 measured |
| LMArena Search | 1166 | Not measured | Only Grok 4.3 measured |
| Neither model has a result on the other 25 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index23
| Benchmark | Grok 4.3 | Hy3 preview | Leader |
|---|---|---|---|
| GPQA Diamond | 91.4% | Not measured | Only Grok 4.3 measured |
| MMMU Pro (Vals AI) | 83.1% | Not measured | Only Grok 4.3 measured |
| GPQA Diamond (Artificial Analysis run) | 90.1% | 86.7% | Grok 4.3 |
| Humanity's Last Exam (Artificial Analysis run) | 37.2% | 27.8% | Grok 4.3 |
| GPQA Diamond (Epoch AI run) | 88.8% | Not measured | Only Grok 4.3 measured |
| Chess puzzles | 18.0% | Not measured | Only Grok 4.3 measured |
| LMArena Text: hard prompts | 1396 | 1413 | Hy3 preview |
| LMArena Text: expert | 1384 | 1422 | Hy3 preview |
| LMArena Text: hard prompts in English | 1402 | 1416 | Hy3 preview |
| LMArena Vision: diagrams | 1241 | Not measured | Only Grok 4.3 measured |
| LMArena Vision: homework | 1241 | Not measured | Only Grok 4.3 measured |
| Neither model has a result on the other 11 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Grok 4.3 | Hy3 preview | Leader |
|---|---|---|---|
| OTIS Mock AIME | 93.3% | Not measured | Only Grok 4.3 measured |
| FrontierMath, tiers 1 to 3 | 42.8% | Not measured | Only Grok 4.3 measured |
| FrontierMath, tier 4 | 14.6% | Not measured | Only Grok 4.3 measured |
| ProofBench | 11.0% | Not measured | Only Grok 4.3 measured |
| LMArena Text: math | 1387 | 1424 | Hy3 preview |
| LMArena Text: mathematical work | 1396 | 1442 | Hy3 preview |
| Neither model has a result on the other 7 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index04
| Benchmark | Grok 4.3 | Hy3 preview | Leader |
|---|---|---|---|
| MMLU Pro | 85.8% | Not measured | Only Grok 4.3 measured |
| LegalBench (Vals AI) | 84.5% | Not measured | Only Grok 4.3 measured |
| CaseLaw (Vals AI) | 79.3% | Not measured | Only Grok 4.3 measured |
| CorpFin (Vals AI) | 68.5% | Not measured | Only Grok 4.3 measured |
| TaxEval (Vals AI) | 70.8% | Not measured | Only Grok 4.3 measured |
| MortgageTax (Vals AI) | 48.3% | Not measured | Only Grok 4.3 measured |
| MedCode (Vals AI) | 38.1% | Not measured | Only Grok 4.3 measured |
| SimpleQA Verified | 33.2% | Not measured | Only Grok 4.3 measured |
| LMArena Text: science | 1399 | 1416 | Hy3 preview |
| LMArena Text: medicine and healthcare | 1384 | 1405 | Hy3 preview |
| LMArena Text: legal and government | 1386 | 1430 | Hy3 preview |
| LMArena Text: business and finance | 1381 | 1397 | Hy3 preview |
| Neither model has a result on the other 6 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
| Benchmark | Grok 4.3 | Hy3 preview | Leader |
|---|---|---|---|
| Design Arena: websites | 1196 | Not measured | Only Grok 4.3 measured |
| Design Arena: UI components | 1192 | Not measured | Only Grok 4.3 measured |
| Design Arena: game dev | 1174 | Not measured | Only Grok 4.3 measured |
| Design Arena: data viz | 1179 | Not measured | Only Grok 4.3 measured |
| Design Arena: 3D | 1125 | Not measured | Only Grok 4.3 measured |
| Design Arena: all code categories | 1187 | Not measured | Only Grok 4.3 measured |
| Design Arena: SVG | 1081 | Not measured | Only Grok 4.3 measured |
| Design Arena: ASCII art | 1125 | Not measured | Only Grok 4.3 measured |
| Neither model has a result on the other 1 board here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index69
| Benchmark | Grok 4.3 | Hy3 preview | Leader |
|---|---|---|---|
| Public Benefits Bench (Vals AI) | 51.7% | Not measured | Only Grok 4.3 measured |
| SAGE (Vals AI) | 19.7% | Not measured | Only Grok 4.3 measured |
| MedScribe (Vals AI) | 74.4% | Not measured | Only Grok 4.3 measured |
| Vals Multimodal Index | 43.3% | Not measured | Only Grok 4.3 measured |
| IFBench | 83.3% | 63.1% | Grok 4.3 |
| Long context reasoning (AA-LCR) | 75.0% | 64.7% | Grok 4.3 |
| DTBench | 90.7% | Not measured | Only Grok 4.3 measured |
| LMCA | 38.3% | Not measured | Only Grok 4.3 measured |
| LMArena Text | 1397 | 1401 | Hy3 preview |
| LMArena Text: instruction following | 1366 | 1383 | Hy3 preview |
| LMArena Text: creative writing | 1380 | 1339 | Grok 4.3 |
| LMArena Vision | 1229 | Not measured | Only Grok 4.3 measured |
| LMArena Text: English | 1404 | 1409 | Hy3 preview |
| LMArena Text: non-English | 1385 | 1390 | Hy3 preview |
| LMArena Text: Chinese | 1422 | 1435 | Hy3 preview |
| LMArena Text: German | 1395 | Not measured | Only Grok 4.3 measured |
| LMArena Text: Japanese | 1379 | Not measured | Only Grok 4.3 measured |
| LMArena Text: Korean | 1354 | Not measured | Only Grok 4.3 measured |
| LMArena Text: French | 1411 | 1425 | Hy3 preview |
| LMArena Text: Spanish | 1399 | 1399 | Grok 4.3 |
| LMArena Text: Russian | 1399 | 1399 | Hy3 preview |
| LMArena Text: Polish | 1398 | Not measured | Only Grok 4.3 measured |
| LMArena Text: multi-turn | 1406 | 1411 | Hy3 preview |
| LMArena Text: longer queries | 1393 | 1402 | Hy3 preview |
| LMArena Text: writing and language | 1385 | 1370 | Grok 4.3 |
| LMArena Text: entertainment and media | 1370 | 1352 | Grok 4.3 |
| LMArena Vision: English | 1235 | Not measured | Only Grok 4.3 measured |
| LMArena Vision: Chinese | 1257 | Not measured | Only Grok 4.3 measured |
| LMArena Vision: reading text | 1235 | Not measured | Only Grok 4.3 measured |
| LMArena Vision: humor | 1222 | Not measured | Only Grok 4.3 measured |
| LMArena Vision: recognizing things | 1217 | Not measured | Only Grok 4.3 measured |
| LMArena Vision: creative writing | 1237 | Not measured | Only Grok 4.3 measured |
| Neither model has a result on the other 14 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Grok 4.3 | Hy3 preview | 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 preview costs less
Repository review
50K input + 3K output tokens
Hy3 preview costs less
Agent loop
200K input + 10K output tokens
Hy3 preview 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 | SpaceXAI Grok 4.3 | Tencent Hy3 preview |
|---|---|---|
| Maker | xAI | Tencent |
| Weights | No public weights listed | Open weights |
| Released | April 30, 2026 | April 22, 2026 |
| Context window | 1M | 262K |
| Input $/Mtok | $1.25 | $0.18 |
| Output $/Mtok | $2.50 | $0.60 |
| Coding index | 37.4 | Not benchmarked |
| Independent measurements behind it | 3 | 1 |
| Cost per solved task | $0.31 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Grok 4.3 better than Hy3 preview for coding?
Only Grok 4.3 has a coding score, 37.4 on the Vibe Coding Index; Hy3 preview has none, so the two cannot be ranked on coding. Pick Hy3 preview for output tokens 4.2x cheaper ($0.60 vs $2.50 per million) and work measured by AA Coding Index or LMArena Text: coding, where it leads.
Which is cheaper, Grok 4.3 or Hy3 preview?
Grok 4.3 costs $1.25 per million input tokens and $2.50 per million output tokens; Hy3 preview costs $0.18 and $0.60. Cost per solved task is known for Grok 4.3 only.
Which has the bigger context window, Grok 4.3 or Hy3 preview?
Grok 4.3, with 1M tokens against 262K.
How many benchmarks is the Grok 4.3 vs Hy3 preview comparison based on?
36 benchmarks have a result for both models as of October 21, 2026. Grok 4.3's coding index rests on 3 independent benchmarks and Hy3 preview's on 1. Where only one model has a result, the page says so instead of naming a winner.