Model comparison
Grok 4.5 vs Llama 3-8B
Grok 4.5 is the stronger model overall, scoring 55.0 to 25.3 on the Noometry Index. Llama 3-8B costs 21× less per token, which makes it the better buy when Grok 4.5's lead doesn't matter for your workload.
Last verified . 22 shared benchmarks.
MetaSummary
- They share 22 benchmarks with published results for both. Grok 4.5 scores higher in 8 categories and Llama 3-8B in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Grok 4.5 leads 62.3 to 7.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 97.8% for Grok 4.5 and 1.9% for Llama 3-8B.
- Llama 3-8B is cheaper at $0.14 / $0.14 per million input/output tokens, against $2 / $6 for Grok 4.5.
- Grok 4.5 accepts more context: 500K tokens versus 8K.
- Llama 3-8B has downloadable open weights; the other is API-only.
Side by side
| Grok 4.5 | Llama 3-8B | |
|---|---|---|
| Provider | xAI | Meta |
| Noometry Index | 55.0 | 25.3 |
| Released | 2026-07-08 | 2024-04-18 |
| Weights | Proprietary | Open |
| Context window | 500K | 8K |
| Max output | 500K | 8K |
| Input $ / M tokens | $2 | $0.14 |
| Output $ / M tokens | $6 | $0.14 |
| Results tracked | 52 | 34 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Grok 4.5 leads
Grok 4.5: 51.7 (#34), Llama 3-8B: 30.5 (#285)
| Benchmark | Grok 4.5 | Llama 3-8B |
|---|---|---|
| LMArena Coding | 1473 | 1152 |
| DeepSWE | 53.8% | — |
| FrontierCode | 42.4% | — |
| LMArena WebDev | 1552 | — |
| SciCode | 54.1% | — |
| WeirdML | 46.4% | — |
| BigCodeBench Instruct | — | 31.9% |
| BigCodeBench Complete | — | 36.9% |
| ALE-Bench | 1,309 | — |
| HumanEval+ | — | 56.7% |
| MBPP+ | — | 54.8% |
Agentic & Tool Use Not comparable
Grok 4.5: 44.4 (#15), Llama 3-8B: —
| Benchmark | Grok 4.5 | Llama 3-8B |
|---|---|---|
| APEX-Agents | 56.2% | — |
| τ²-bench Banking | 47.9% | — |
| PostTrainBench | 23.4% | — |
| GBAEval | 65.4% | — |
| GDP.pdf | 14% | — |
| LMArena Search | 1213 | — |
| Vending-Bench 2 | 3,887 | — |
Reasoning Grok 4.5 leads
Grok 4.5: 56.0 (#23), Llama 3-8B: 13.8 (#321)
| Benchmark | Grok 4.5 | Llama 3-8B |
|---|---|---|
| Chess Puzzles | 36% | 0% |
| LMArena Hard Prompts | 1462 | 1133 |
| DTBench | 96.5% | 43.9% |
| Epoch Capabilities Index | 154.02 | 116.33 |
| ARC-AGI-2 | 52.6% | — |
| SimpleBench | 70% | — |
| Kagi LLM Benchmark | 83.5% | — |
| NYT Connections (extended) | 79.9% | — |
| ARC-AGI-1 | 87.2% | — |
| CritPt | 15.4% | — |
| LMCA | 45.2% | — |
| Surface Evolver Bench | 74.4% | — |
| Adversarial NLI | — | 57.3% |
| ForecastBench | — | 58.6 |
| WinoGrande | — | 75.7% |
Math Grok 4.5 leads
Grok 4.5: 60.7 (#29), Llama 3-8B: 8.3 (#317)
| Benchmark | Grok 4.5 | Llama 3-8B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 97.8% | 1.9% |
| LMArena Math | 1457 | 1151 |
| FrontierMath (Tiers 1-3) | 57.2% | — |
| FrontierMath Tier 4 | 24.4% | — |
| ProofBench | 31% | — |
| MATH Level 5 | — | 6.1% |
Knowledge Grok 4.5 leads
Grok 4.5: 62.3 (#21), Llama 3-8B: 7.8 (#303)
| Benchmark | Grok 4.5 | Llama 3-8B |
|---|---|---|
| GPQA Diamond | 93.4% | 26.1% |
| LMArena Expert | 1466 | 1113 |
| SimpleQA Verified | 48.3% | — |
| ARC (AI2) Challenge | — | 82.8% |
| MMLU | — | 68.8% |
| OpenBookQA | — | 82.6% |
| TriviaQA | — | 67.7% |
Multimodal Not comparable
Grok 4.5: 37.8 (#63), Llama 3-8B: —
| Benchmark | Grok 4.5 | Llama 3-8B |
|---|---|---|
| LMArena Vision | 1287 | — |
| Blueprint-Bench 2 | 27.3% | — |
| Furniture Assembly | 22.5% | — |
| LMArena Document | 1452 | — |
Multilingual Grok 4.5 leads
Grok 4.5: 54.5 (#43), Llama 3-8B: 31.0 (#256)
| Benchmark | Grok 4.5 | Llama 3-8B |
|---|---|---|
| LMArena Non-English | 1439 | 1098 |
| LMArena Chinese | 1497 | 1076 |
| LMArena French | 1458 | 1158 |
| LMArena German | 1448 | 1104 |
| LMArena Japanese | 1425 | 967 |
| LMArena Korean | 1402 | 1003 |
| LMArena Russian | 1447 | 1109 |
| LMArena Spanish | 1452 | 1172 |
Instruction Following Grok 4.5 leads
Grok 4.5: 76.3 (#45), Llama 3-8B: 58.7 (#255)
| Benchmark | Grok 4.5 | Llama 3-8B |
|---|---|---|
| LMArena Instruction Following | 1446 | 1127 |
Long Context Grok 4.5 leads
Grok 4.5: 44.8 (#54), Llama 3-8B: 34.2 (#246)
| Benchmark | Grok 4.5 | Llama 3-8B |
|---|---|---|
| LMArena Longer Query | 1463 | 1128 |
Writing & Preference Grok 4.5 leads
Grok 4.5: 65.9 (#40), Llama 3-8B: 37.7 (#251)
| Benchmark | Grok 4.5 | Llama 3-8B |
|---|---|---|
| LMArena Text | 1448 | 1166 |
| LMArena Creative Writing | 1442 | 1150 |
| LMArena Multi-Turn | 1457 | 1152 |
| EQ-Bench Creative Writing | 1579 | — |
Frequently asked questions
Is Grok 4.5 better than Llama 3-8B?
Grok 4.5 is the stronger model overall, scoring 55.0 to 25.3 on the Noometry Index. Llama 3-8B costs 21× less per token, which makes it the better buy when Grok 4.5's lead doesn't matter for your workload.
Which is cheaper, Grok 4.5 or Llama 3-8B?
Llama 3-8B is cheaper. It lists at $0.14 per million input tokens and $0.14 per million output tokens; Grok 4.5 lists at $2 and $6.
Is Grok 4.5 or Llama 3-8B better for coding?
Grok 4.5 scores higher on coding benchmarks: 51.7 versus 30.5 in the Noometry coding category.
Which has the bigger context window?
Grok 4.5 does, with 500K tokens against 8K.
How many benchmarks do Grok 4.5 and Llama 3-8B share?
22 benchmarks have published results for both models. Grok 4.5 has 52 scored results on Noometry and Llama 3-8B has 34.