# GPT-5.5 vs Llama 3-8B

> GPT-5.5 is the stronger model overall, scoring 63.4 to 25.3 on the Noometry Index. Llama 3-8B costs 80× less per token, which makes it the better buy when GPT-5.5's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/gpt-5-5-vs-llama-3-8b
- Last updated: 2026-10-10
- Shared benchmarks: 23

## Summary

- They share 23 benchmarks with published results for both. GPT-5.5 scores higher in 8 categories and Llama 3-8B in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.5 leads 81.5 to 8.3.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 100% for GPT-5.5 and 1.9% for Llama 3-8B.
- Llama 3-8B is cheaper at $0.14 / $0.14 per million input/output tokens, against $5 / $30 for GPT-5.5.
- GPT-5.5 accepts more context: 1.05M tokens versus 8K.
- Llama 3-8B has downloadable open weights; the other is API-only.

## Snapshot

| | GPT-5.5 | Llama 3-8B |
|---|---|---|
| Provider | OpenAI | Meta |
| Noometry Index | 63.4 | 25.3 |
| Rank | 8 | 338 |
| Context | 1.05M | 8K |
| Input $/M | $5 | $0.14 |
| Output $/M | $30 | $0.14 |
| Weights | Proprietary | Open |

## Coding

- GPT-5.5: 58.3 (#13)
- Llama 3-8B: 30.5 (#285)

| Benchmark | GPT-5.5 | Llama 3-8B |
|---|---|---|
| LMArena Coding | 1494 | 1152 |
| SWE-bench Verified | 80.6% | — |
| DeepSWE | 67% | — |
| FrontierCode | 43% | — |
| LMArena WebDev | 1512 | — |
| SciCode | 56.1% | — |
| GSO | 40.2% | — |
| WeirdML | 84.9% | — |
| BigCodeBench Instruct | — | 31.9% |
| MirrorCode | 10% | — |
| BigCodeBench Complete | — | 36.9% |
| ALE-Bench | 1,943 | — |
| HumanEval+ | — | 56.7% |
| MBPP+ | — | 54.8% |

## Agentic & Tool Use

- GPT-5.5: 50.7 (#6)
- Llama 3-8B: —

| Benchmark | GPT-5.5 | Llama 3-8B |
|---|---|---|
| Terminal-Bench | 84.7% | — |
| APEX-Agents | 55.1% | — |
| OSWorld 2.0 | 13% | — |
| Remote Labor Index | 6.3% | — |
| τ²-bench Banking | 44.6% | — |
| DeepResearch Bench | 54% | — |
| PostTrainBench | 27.2% | — |
| ExploitBench | 47.4% | — |
| GBAEval | 53.2% | — |
| GDP.pdf | 26% | — |
| LMArena Search | 1242 | — |
| Vending-Bench 2 | 7,524 | — |

## Reasoning

- GPT-5.5: 72.7 (#8)
- Llama 3-8B: 13.8 (#321)

| Benchmark | GPT-5.5 | Llama 3-8B |
|---|---|---|
| Chess Puzzles | 54% | 0% |
| LMArena Hard Prompts | 1489 | 1133 |
| DTBench | 96% | 43.9% |
| Epoch Capabilities Index | 159.25 | 116.33 |
| ForecastBench | 60.6 | 58.6 |
| ARC-AGI-2 | 85% | — |
| SimpleBench | 69% | — |
| Kagi LLM Benchmark | 88.8% | — |
| NYT Connections (extended) | 96.2% | — |
| ARC-AGI-1 | 95% | — |
| CritPt | 27.1% | — |
| EBR-Bench | 34.3% | — |
| Mystery Game Puzzles | 56% | — |
| LMCA | 54.3% | — |
| Surface Evolver Bench | 88.1% | — |
| Adversarial NLI | — | 57.3% |
| Bench to the Future 3 | 0.14 | — |
| WinoGrande | — | 75.7% |

## Math

- GPT-5.5: 81.5 (#9)
- Llama 3-8B: 8.3 (#317)

| Benchmark | GPT-5.5 | Llama 3-8B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 100% | 1.9% |
| LMArena Math | 1486 | 1151 |
| FrontierMath (Tiers 1-3) | 85.3% | — |
| FrontierMath Tier 4 | 72.5% | — |
| MathArena Final-Answer Competitions | 94.3% | — |
| ProofBench | 50% | — |
| MATH Level 5 | — | 6.1% |
| FrontierMath (Feb 2025 set) | 51.7% | — |
| FrontierMath Erdős | 0% | — |
| FrontierMath Tier 4 (v1) | 35.4% | — |

## Knowledge

- GPT-5.5: 64.4 (#14)
- Llama 3-8B: 7.8 (#303)

| Benchmark | GPT-5.5 | Llama 3-8B |
|---|---|---|
| GPQA Diamond | 94% | 26.1% |
| LMArena Expert | 1508 | 1113 |
| SimpleQA Verified | 63% | — |
| Vectara Hallucination Rate | 9.3% | — |
| ARC (AI2) Challenge | — | 82.8% |
| MMLU | — | 68.8% |
| OpenBookQA | — | 82.6% |
| TriviaQA | — | 67.7% |

## Multimodal

- GPT-5.5: 47.1 (#11)
- Llama 3-8B: —

| Benchmark | GPT-5.5 | Llama 3-8B |
|---|---|---|
| LMArena Vision | 1297 | — |
| Blueprint-Bench 2 | 36.2% | — |
| Furniture Assembly | 44.2% | — |
| LMArena Document | 1486 | — |

## Multilingual

- GPT-5.5: 56.5 (#19)
- Llama 3-8B: 31.0 (#256)

| Benchmark | GPT-5.5 | Llama 3-8B |
|---|---|---|
| LMArena Non-English | 1467 | 1098 |
| LMArena Chinese | 1535 | 1076 |
| LMArena French | 1486 | 1158 |
| LMArena German | 1478 | 1104 |
| LMArena Japanese | 1498 | 967 |
| LMArena Korean | 1461 | 1003 |
| LMArena Russian | 1473 | 1109 |
| LMArena Spanish | 1469 | 1172 |

## Instruction Following

- GPT-5.5: 77.7 (#18)
- Llama 3-8B: 58.7 (#255)

| Benchmark | GPT-5.5 | Llama 3-8B |
|---|---|---|
| LMArena Instruction Following | 1479 | 1127 |

## Long Context

- GPT-5.5: 48.3 (#12)
- Llama 3-8B: 34.2 (#246)

| Benchmark | GPT-5.5 | Llama 3-8B |
|---|---|---|
| LMArena Longer Query | 1484 | 1128 |
| CL-bench Life | 22.2% | — |

## Writing & Preference

- GPT-5.5: 72.8 (#13)
- Llama 3-8B: 37.7 (#251)

| Benchmark | GPT-5.5 | Llama 3-8B |
|---|---|---|
| LMArena Text | 1471 | 1166 |
| LMArena Creative Writing | 1455 | 1150 |
| LMArena Multi-Turn | 1477 | 1152 |
| EQ-Bench Creative Writing | 1844 | — |
| EQ-Bench 4 | 1315 | — |

## FAQ

### Is GPT-5.5 better than Llama 3-8B?

GPT-5.5 is the stronger model overall, scoring 63.4 to 25.3 on the Noometry Index. Llama 3-8B costs 80× less per token, which makes it the better buy when GPT-5.5's lead doesn't matter for your workload.

### Which is cheaper, GPT-5.5 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; GPT-5.5 lists at $5 and $30.

### Is GPT-5.5 or Llama 3-8B better for coding?

GPT-5.5 scores higher on coding benchmarks: 58.3 versus 30.5 in the Noometry coding category.

### Which has the bigger context window?

GPT-5.5 does, with 1.05M tokens against 8K.

### How many benchmarks do GPT-5.5 and Llama 3-8B share?

23 benchmarks have published results for both models. GPT-5.5 has 71 scored results on Noometry and Llama 3-8B has 34.