Model comparison
GLM-4.5-Air vs Llama-3.3-70B-Instruct
GLM-4.5-Air is the stronger model overall, scoring 38.9 to 30.6 on the Noometry Index. Llama-3.3-70B-Instruct costs 2.7× less per token, which makes it the better buy when GLM-4.5-Air's lead doesn't matter for your workload.
Last verified . 19 shared benchmarks.
MetaSummary
- They share 19 benchmarks with published results for both. GLM-4.5-Air scores higher in 7 categories and Llama-3.3-70B-Instruct in 1 category; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where GLM-4.5-Air leads 36.2 to 15.3.
- The biggest single-benchmark swing is Vectara Hallucination Rate: 9.3% for GLM-4.5-Air and 4.1% for Llama-3.3-70B-Instruct.
- Llama-3.3-70B-Instruct is cheaper at $0.10 / $0.32 per million input/output tokens, against $0.20 / $1.10 for GLM-4.5-Air.
- GLM-4.5-Air accepts more context: 131K tokens versus 128K.
Side by side
| GLM-4.5-Air | Llama-3.3-70B-Instruct | |
|---|---|---|
| Provider | Z.ai (Zhipu) | Meta |
| Noometry Index | 38.9 | 30.6 |
| Released | 2025-07-20 | 2024-12-06 |
| Weights | Open | Open |
| Context window | 131K | 128K |
| Max output | 98K | 4K |
| Input $ / M tokens | $0.20 | $0.10 |
| Output $ / M tokens | $1.10 | $0.32 |
| Results tracked | 27 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GLM-4.5-Air leads
GLM-4.5-Air: 33.3 (#259), Llama-3.3-70B-Instruct: 31.0 (#290)
| Benchmark | GLM-4.5-Air | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Coding | 1397 | 1268 |
| SciCode | — | 26% |
| GSO | 2.9% | — |
| WeirdML | — | 14.4% |
| BigCodeBench Instruct | — | 46.9% |
| LiveBench Coding | — | 36.6% |
| BigCodeBench Complete | — | 57.5% |
Agentic & Tool Use Not comparable
GLM-4.5-Air: —, Llama-3.3-70B-Instruct: 25.8 (#105)
| Benchmark | GLM-4.5-Air | Llama-3.3-70B-Instruct |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 31.9% |
| BALROG | — | 23% |
Reasoning GLM-4.5-Air leads
GLM-4.5-Air: 24.1 (#166), Llama-3.3-70B-Instruct: 14.1 (#327)
| Benchmark | GLM-4.5-Air | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Hard Prompts | 1379 | 1257 |
| ForecastBench | 59.2 | 58.6 |
| SimpleBench | — | 19.9% |
| Kagi LLM Benchmark | 43% | — |
| CritPt | — | 0% |
| LiveBench Reasoning | — | 50.8% |
| DTBench | — | 59.5% |
| LiveBench Data Analysis | — | 49.5% |
| LMCA | — | 17.5% |
| Epoch Capabilities Index | — | 127.33 |
| LiveBench | — | 50.2% |
Math GLM-4.5-Air leads
GLM-4.5-Air: 36.2 (#170), Llama-3.3-70B-Instruct: 15.3 (#298)
| Benchmark | GLM-4.5-Air | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Math | 1396 | 1267 |
| OTIS Mock AIME 2024-2025 | — | 5.1% |
| Omni-MATH | 39.1% | — |
| LiveBench Math | — | 42.2% |
| MATH Level 5 | — | 41.6% |
Knowledge GLM-4.5-Air leads
GLM-4.5-Air: 35.0 (#191), Llama-3.3-70B-Instruct: 30.6 (#226)
| Benchmark | GLM-4.5-Air | Llama-3.3-70B-Instruct |
|---|---|---|
| Vectara Hallucination Rate | 9.3% | 4.1% |
| LMArena Expert | 1370 | 1225 |
| GPQA Diamond | — | 47.4% |
| Humanity's Last Exam | 8.1% | — |
| MMLU-Pro | 76.2% | — |
| Confabulations | — | 22.8% |
| GPQA (HELM) | 59.4% | — |
| MMLU | — | 86.3% |
Multilingual GLM-4.5-Air leads
GLM-4.5-Air: 49.1 (#135), Llama-3.3-70B-Instruct: 39.9 (#220)
| Benchmark | GLM-4.5-Air | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Non-English | 1366 | 1236 |
| LMArena Chinese | 1426 | 1217 |
| LMArena French | 1399 | 1281 |
| LMArena German | 1377 | 1251 |
| LMArena Japanese | 1348 | 1150 |
| LMArena Korean | 1308 | 1143 |
| LMArena Russian | 1373 | 1252 |
| LMArena Spanish | 1386 | 1270 |
Instruction Following Llama-3.3-70B-Instruct leads
GLM-4.5-Air: 69.6 (#171), Llama-3.3-70B-Instruct: 71.1 (#157)
| Benchmark | GLM-4.5-Air | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Instruction Following | 1354 | 1242 |
| LiveBench Instruction Following | — | 82.7% |
| IFEval | 81.2% | — |
Long Context GLM-4.5-Air leads
GLM-4.5-Air: 41.6 (#135), Llama-3.3-70B-Instruct: 26.4 (#295)
| Benchmark | GLM-4.5-Air | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Longer Query | 1366 | 1256 |
| Fiction.LiveBench | — | 33.3% |
Writing & Preference GLM-4.5-Air leads
GLM-4.5-Air: 55.9 (#139), Llama-3.3-70B-Instruct: 47.6 (#207)
| Benchmark | GLM-4.5-Air | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Text | 1384 | 1274 |
| LMArena Creative Writing | 1343 | 1250 |
| LMArena Multi-Turn | 1371 | 1280 |
| WildBench | 78.9% | — |
| LiveBench Language | — | 39.2% |
Frequently asked questions
Is GLM-4.5-Air better than Llama-3.3-70B-Instruct?
GLM-4.5-Air is the stronger model overall, scoring 38.9 to 30.6 on the Noometry Index. Llama-3.3-70B-Instruct costs 2.7× less per token, which makes it the better buy when GLM-4.5-Air's lead doesn't matter for your workload.
Which is cheaper, GLM-4.5-Air or Llama-3.3-70B-Instruct?
Llama-3.3-70B-Instruct is cheaper. It lists at $0.10 per million input tokens and $0.32 per million output tokens; GLM-4.5-Air lists at $0.20 and $1.10.
Is GLM-4.5-Air or Llama-3.3-70B-Instruct better for coding?
GLM-4.5-Air scores higher on coding benchmarks: 33.3 versus 31.0 in the Noometry coding category.
Which has the bigger context window?
GLM-4.5-Air does, with 131K tokens against 128K.
How many benchmarks do GLM-4.5-Air and Llama-3.3-70B-Instruct share?
19 benchmarks have published results for both models. GLM-4.5-Air has 27 scored results on Noometry and Llama-3.3-70B-Instruct has 43.