Model comparison
Gemma 4 31B IT vs Llama-3.3-70B-Instruct
Gemma 4 31B IT is the stronger model overall, scoring 43.2 to 30.3 on the Noometry Index.
Last verified . 23 shared benchmarks.
Google
MetaSummary
- They share 23 benchmarks with published results for both. Gemma 4 31B IT 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 Gemma 4 31B IT leads 41.3 to 14.3.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 73.3% for Gemma 4 31B IT and 5.1% for Llama-3.3-70B-Instruct.
- Both cost about the same: $0.09 input and $0.34 output per million tokens.
- Gemma 4 31B IT accepts more context: 262K tokens versus 128K.
Side by side
| Gemma 4 31B IT | Llama-3.3-70B-Instruct | |
|---|---|---|
| Provider | Meta | |
| Noometry Index | 43.2 | 30.3 |
| Released | 2026-04-02 | 2024-12-06 |
| Weights | Open | Open |
| Context window | 262K | 128K |
| Max output | 33K | 4K |
| Input $ / M tokens | $0.09 | $0.10 |
| Output $ / M tokens | $0.34 | $0.32 |
| Results tracked | 35 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Gemma 4 31B IT leads
Gemma 4 31B IT: 41.9 (#98), Llama-3.3-70B-Instruct: 30.6 (#284)
| Benchmark | Gemma 4 31B IT | Llama-3.3-70B-Instruct |
|---|---|---|
| SciCode | 43.4% | 26% |
| WeirdML | 52.3% | 14.4% |
| LMArena Coding | 1459 | 1268 |
| LMArena WebDev | 1364 | — |
| BigCodeBench Instruct | — | 46.9% |
| LiveBench Coding | — | 36.6% |
| BigCodeBench Complete | — | 57.5% |
| ALE-Bench | 925.5 | — |
Agentic & Tool Use Not comparable
Gemma 4 31B IT: —, Llama-3.3-70B-Instruct: 25.6 (#100)
| Benchmark | Gemma 4 31B IT | Llama-3.3-70B-Instruct |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 31.9% |
| BALROG | — | 23% |
Reasoning Gemma 4 31B IT leads
Gemma 4 31B IT: 26.4 (#119), Llama-3.3-70B-Instruct: 13.7 (#325)
| Benchmark | Gemma 4 31B IT | Llama-3.3-70B-Instruct |
|---|---|---|
| CritPt | 1.4% | 0% |
| LMArena Hard Prompts | 1448 | 1257 |
| DTBench | 82.7% | 59.5% |
| LMCA | 39.3% | 17.5% |
| Epoch Capabilities Index | 142.74 | 127.32 |
| SimpleBench | — | 19.9% |
| Kagi LLM Benchmark | 63.5% | — |
| NYT Connections (extended) | 70.6% | — |
| Chess Puzzles | 5% | — |
| Thematic Generalization | 53% | — |
| LiveBench Reasoning | — | 50.8% |
| LiveBench Data Analysis | — | 49.5% |
| Surface Evolver Bench | 30.6% | — |
| ForecastBench | — | 58.6 |
| LiveBench | — | 50.2% |
Math Gemma 4 31B IT leads
Gemma 4 31B IT: 41.3 (#72), Llama-3.3-70B-Instruct: 14.3 (#292)
| Benchmark | Gemma 4 31B IT | Llama-3.3-70B-Instruct |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 73.3% | 5.1% |
| LMArena Math | 1464 | 1268 |
| LiveBench Math | — | 42.2% |
| MATH Level 5 | — | 41.6% |
Knowledge Gemma 4 31B IT leads
Gemma 4 31B IT: 37.9 (#141), Llama-3.3-70B-Instruct: 30.5 (#220)
| Benchmark | Gemma 4 31B IT | Llama-3.3-70B-Instruct |
|---|---|---|
| GPQA Diamond | 75.8% | 47.4% |
| Vectara Hallucination Rate | 7.4% | 4.1% |
| LMArena Expert | 1465 | 1225 |
| SimpleQA Verified | 10.4% | — |
| Confabulations | — | 22.8% |
| MMLU | — | 86.3% |
Multimodal Not comparable
Gemma 4 31B IT: 41.0 (#32), Llama-3.3-70B-Instruct: —
| Benchmark | Gemma 4 31B IT | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Vision | 1278 | — |
| LMArena Document | 1425 | — |
Multilingual Gemma 4 31B IT leads
Gemma 4 31B IT: 54.0 (#52), Llama-3.3-70B-Instruct: 40.1 (#215)
| Benchmark | Gemma 4 31B IT | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Non-English | 1431 | 1236 |
| LMArena Chinese | 1478 | 1217 |
| LMArena French | 1436 | 1280 |
| LMArena Russian | 1462 | 1252 |
| LMArena Spanish | 1443 | 1269 |
| LMArena German | — | 1251 |
| LMArena Japanese | — | 1149 |
| LMArena Korean | — | 1142 |
Instruction Following Gemma 4 31B IT leads
Gemma 4 31B IT: 75.6 (#58), Llama-3.3-70B-Instruct: 71.3 (#152)
| Benchmark | Gemma 4 31B IT | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Instruction Following | 1433 | 1242 |
| LiveBench Instruction Following | — | 82.7% |
Long Context Gemma 4 31B IT leads
Gemma 4 31B IT: 44.2 (#67), Llama-3.3-70B-Instruct: 26.4 (#290)
| Benchmark | Gemma 4 31B IT | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Longer Query | 1446 | 1256 |
| Fiction.LiveBench | — | 33.3% |
Writing & Preference Gemma 4 31B IT leads
Gemma 4 31B IT: 60.7 (#92), Llama-3.3-70B-Instruct: 47.8 (#203)
| Benchmark | Gemma 4 31B IT | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Text | 1443 | 1274 |
| LMArena Creative Writing | 1416 | 1250 |
| LMArena Multi-Turn | 1452 | 1279 |
| EQ-Bench Creative Writing | 1368 | — |
| EQ-Bench 4 | 1120 | — |
| LiveBench Language | — | 39.2% |
Frequently asked questions
Is Gemma 4 31B IT better than Llama-3.3-70B-Instruct?
Gemma 4 31B IT is the stronger model overall, scoring 43.2 to 30.3 on the Noometry Index.
Which is cheaper, Gemma 4 31B IT or Llama-3.3-70B-Instruct?
Gemma 4 31B IT is cheaper. It lists at $0.09 per million input tokens and $0.34 per million output tokens; Llama-3.3-70B-Instruct lists at $0.10 and $0.32.
Is Gemma 4 31B IT or Llama-3.3-70B-Instruct better for coding?
Gemma 4 31B IT scores higher on coding benchmarks: 41.9 versus 30.6 in the Noometry coding category.
Which has the bigger context window?
Gemma 4 31B IT does, with 262K tokens against 128K.
How many benchmarks do Gemma 4 31B IT and Llama-3.3-70B-Instruct share?
23 benchmarks have published results for both models. Gemma 4 31B IT has 35 scored results on Noometry and Llama-3.3-70B-Instruct has 43.