Model comparison
Hy3 vs Llama-3.3-70B-Instruct
Hy3 is the stronger model overall, scoring 44.2 to 30.6 on the Noometry Index.
Last verified . 17 shared benchmarks.
Tencent
MetaSummary
- They share 17 benchmarks with published results for both. Hy3 scores higher in 8 categories and Llama-3.3-70B-Instruct in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where Hy3 leads 40.1 to 15.3.
- Hy3 is cheaper at $0.0825 / $0.33 per million input/output tokens, against $0.10 / $0.32 for Llama-3.3-70B-Instruct.
- Hy3 accepts more context: 262K tokens versus 128K.
Side by side
| Hy3 | Llama-3.3-70B-Instruct | |
|---|---|---|
| Provider | Tencent | Meta |
| Noometry Index | 44.2 | 30.6 |
| Released | 2026-07-06 | 2024-12-06 |
| Weights | Open | Open |
| Context window | 262K | 128K |
| Max output | 128K | 4K |
| Input $ / M tokens | $0.0825 | $0.10 |
| Output $ / M tokens | $0.33 | $0.32 |
| Results tracked | 19 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Hy3 leads
Hy3: 46.8 (#63), Llama-3.3-70B-Instruct: 31.0 (#290)
| Benchmark | Hy3 | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Coding | 1464 | 1268 |
| LMArena WebDev | 1508 | — |
| SciCode | — | 26% |
| WeirdML | — | 14.4% |
| BigCodeBench Instruct | — | 46.9% |
| LiveBench Coding | — | 36.6% |
| BigCodeBench Complete | — | 57.5% |
Agentic & Tool Use Not comparable
Hy3: —, Llama-3.3-70B-Instruct: 25.8 (#105)
| Benchmark | Hy3 | Llama-3.3-70B-Instruct |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 31.9% |
| BALROG | — | 23% |
Reasoning Hy3 leads
Hy3: 26.1 (#136), Llama-3.3-70B-Instruct: 14.1 (#327)
| Benchmark | Hy3 | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Hard Prompts | 1447 | 1257 |
| SimpleBench | — | 19.9% |
| NYT Connections (extended) | 41.2% | — |
| CritPt | — | 0% |
| LiveBench Reasoning | — | 50.8% |
| DTBench | — | 59.5% |
| LiveBench Data Analysis | — | 49.5% |
| LMCA | — | 17.5% |
| Epoch Capabilities Index | — | 127.33 |
| ForecastBench | — | 58.6 |
| LiveBench | — | 50.2% |
Math Hy3 leads
Hy3: 40.1 (#93), Llama-3.3-70B-Instruct: 15.3 (#298)
| Benchmark | Hy3 | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Math | 1475 | 1267 |
| OTIS Mock AIME 2024-2025 | — | 5.1% |
| LiveBench Math | — | 42.2% |
| MATH Level 5 | — | 41.6% |
Knowledge Hy3 leads
Hy3: 40.8 (#114), Llama-3.3-70B-Instruct: 30.6 (#226)
| Benchmark | Hy3 | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Expert | 1460 | 1225 |
| GPQA Diamond | — | 47.4% |
| Confabulations | — | 22.8% |
| Vectara Hallucination Rate | — | 4.1% |
| MMLU | — | 86.3% |
Multilingual Hy3 leads
Hy3: 53.5 (#65), Llama-3.3-70B-Instruct: 39.9 (#220)
| Benchmark | Hy3 | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Non-English | 1426 | 1236 |
| LMArena Chinese | 1493 | 1217 |
| LMArena French | 1461 | 1281 |
| LMArena German | 1439 | 1251 |
| LMArena Japanese | 1392 | 1150 |
| LMArena Korean | 1395 | 1143 |
| LMArena Russian | 1432 | 1252 |
| LMArena Spanish | 1456 | 1270 |
Instruction Following Hy3 leads
Hy3: 75.1 (#70), Llama-3.3-70B-Instruct: 71.1 (#157)
| Benchmark | Hy3 | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Instruction Following | 1426 | 1242 |
| LiveBench Instruction Following | — | 82.7% |
Long Context Hy3 leads
Hy3: 44.1 (#75), Llama-3.3-70B-Instruct: 26.4 (#295)
| Benchmark | Hy3 | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Longer Query | 1442 | 1256 |
| Fiction.LiveBench | — | 33.3% |
Writing & Preference Hy3 leads
Hy3: 62.2 (#81), Llama-3.3-70B-Instruct: 47.6 (#207)
| Benchmark | Hy3 | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Text | 1439 | 1274 |
| LMArena Creative Writing | 1402 | 1250 |
| LMArena Multi-Turn | 1436 | 1280 |
| LiveBench Language | — | 39.2% |
Frequently asked questions
Is Hy3 better than Llama-3.3-70B-Instruct?
Hy3 is the stronger model overall, scoring 44.2 to 30.6 on the Noometry Index.
Which is cheaper, Hy3 or Llama-3.3-70B-Instruct?
Hy3 is cheaper. It lists at $0.0825 per million input tokens and $0.33 per million output tokens; Llama-3.3-70B-Instruct lists at $0.10 and $0.32.
Is Hy3 or Llama-3.3-70B-Instruct better for coding?
Hy3 scores higher on coding benchmarks: 46.8 versus 31.0 in the Noometry coding category.
Which has the bigger context window?
Hy3 does, with 262K tokens against 128K.
How many benchmarks do Hy3 and Llama-3.3-70B-Instruct share?
17 benchmarks have published results for both models. Hy3 has 19 scored results on Noometry and Llama-3.3-70B-Instruct has 43.