Model comparison
GPT-5.4 vs Llama 4 Scout
GPT-5.4 is the stronger model overall, scoring 59.5 to 27.4 on the Noometry Index. Llama 4 Scout costs 37× less per token, which makes it the better buy when GPT-5.4's lead doesn't matter for your workload.
Last verified . 32 shared benchmarks.
MetaSummary
- They share 32 benchmarks with published results for both. GPT-5.4 scores higher in 10 categories and Llama 4 Scout in 0 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.4 leads 73.3 to 18.4.
- The biggest single-benchmark swing is ARC-AGI-1: 93.7% for GPT-5.4 and 0.5% for Llama 4 Scout.
- Llama 4 Scout is cheaper at $0.10 / $0.30 per million input/output tokens, against $2.50 / $15 for GPT-5.4.
- GPT-5.4 accepts more context: 1.05M tokens versus 128K.
- Llama 4 Scout has downloadable open weights; the other is API-only.
Side by side
| GPT-5.4 | Llama 4 Scout | |
|---|---|---|
| Provider | OpenAI | Meta |
| Noometry Index | 59.5 | 27.4 |
| Released | 2026-03-05 | 2025-04-05 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 128K |
| Max output | 128K | 4K |
| Input $ / M tokens | $2.50 | $0.10 |
| Output $ / M tokens | $15 | $0.30 |
| Results tracked | 68 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.4 leads
GPT-5.4: 52.8 (#28), Llama 4 Scout: 19.9 (#334)
| Benchmark | GPT-5.4 | Llama 4 Scout |
|---|---|---|
| SciCode | 56.6% | 17% |
| LMArena Coding | 1497 | 1286 |
| SWE-bench Verified | 76.9% | — |
| DeepSWE | 51.8% | — |
| SWE-bench Verified (bash only) | — | 9.1% |
| LMArena WebDev | 1465 | — |
| GSO | 31.4% | — |
| WeirdML | 77.7% | — |
| MirrorCode | 15.6% | — |
| BigCodeBench Complete | — | 43.1% |
| ALE-Bench | 1,607 | — |
| AlgoTune | 1.85 | — |
Agentic & Tool Use GPT-5.4 leads
GPT-5.4: 46.5 (#12), Llama 4 Scout: 24.4 (#109)
| Benchmark | GPT-5.4 | Llama 4 Scout |
|---|---|---|
| Terminal-Bench | 81.8% | — |
| APEX-Agents | 52.4% | — |
| Berkeley Function Calling Leaderboard | — | 28.1% |
| τ²-bench Banking | 39.4% | — |
| DeepResearch Bench | 35.1% | — |
| PostTrainBench | 19% | — |
| GBAEval | 45.1% | — |
| LMArena Search | 1197 | — |
| METR Time Horizons | 74.3% | — |
| Vending-Bench 2 | 6,144 | — |
Reasoning GPT-5.4 leads
GPT-5.4: 61.7 (#17), Llama 4 Scout: 9.2 (#339)
| Benchmark | GPT-5.4 | Llama 4 Scout |
|---|---|---|
| ARC-AGI-2 | 74% | 0% |
| Kagi LLM Benchmark | 63.8% | 36.9% |
| ARC-AGI-1 | 93.7% | 0.5% |
| CritPt | 23.4% | 0% |
| LMArena Hard Prompts | 1485 | 1266 |
| DTBench | 94.4% | 57.9% |
| LMCA | 52% | 12% |
| Epoch Capabilities Index | 156.92 | 129.64 |
| ForecastBench | 59.5 | 57.5 |
| NYT Connections (extended) | 91.3% | — |
| Chess Puzzles | 44% | — |
| EnigmaEval | 16% | — |
| Thematic Generalization | 80% | — |
| EBR-Bench | 25.4% | — |
| Mystery Game Puzzles | 37% | — |
Math GPT-5.4 leads
GPT-5.4: 73.3 (#14), Llama 4 Scout: 18.4 (#280)
| Benchmark | GPT-5.4 | Llama 4 Scout |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 97.8% | 7.8% |
| LMArena Math | 1488 | 1286 |
| FrontierMath (Feb 2025 set) | 47.6% | 0% |
| FrontierMath (Tiers 1-3) | 78.6% | — |
| FrontierMath Tier 4 | 49% | — |
| MathArena Final-Answer Competitions | 83.1% | — |
| ProofBench | 56% | — |
| Omni-MATH | — | 37.3% |
| MATH Level 5 | — | 62.3% |
| FrontierMath Tier 4 (v1) | 27.1% | — |
Knowledge GPT-5.4 leads
GPT-5.4: 65.3 (#13), Llama 4 Scout: 31.8 (#208)
| Benchmark | GPT-5.4 | Llama 4 Scout |
|---|---|---|
| GPQA Diamond | 93.3% | 51.8% |
| Vectara Hallucination Rate | 7% | 7.7% |
| LMArena Expert | 1507 | 1236 |
| Humanity's Last Exam | 36.2% | — |
| SimpleQA Verified | 45.1% | — |
| MMLU-Pro | — | 74.2% |
| GPQA (HELM) | — | 50.7% |
Multimodal GPT-5.4 leads
GPT-5.4: 43.8 (#19), Llama 4 Scout: 31.6 (#98)
| Benchmark | GPT-5.4 | Llama 4 Scout |
|---|---|---|
| LMArena Vision | 1302 | 1118 |
| Blueprint-Bench 2 | 27.1% | — |
| Furniture Assembly | 37.5% | — |
| LMArena Document | 1471 | — |
| SpatialViz-Bench | — | 34.2% |
Multilingual GPT-5.4 leads
GPT-5.4: 56.4 (#22), Llama 4 Scout: 41.3 (#207)
| Benchmark | GPT-5.4 | Llama 4 Scout |
|---|---|---|
| LMArena Non-English | 1465 | 1253 |
| LMArena Chinese | 1520 | 1256 |
| LMArena French | 1493 | 1283 |
| LMArena German | 1472 | 1274 |
| LMArena Japanese | 1484 | 1209 |
| LMArena Korean | 1449 | 1207 |
| LMArena Russian | 1480 | 1263 |
| LMArena Spanish | 1453 | 1277 |
Instruction Following GPT-5.4 leads
GPT-5.4: 77.3 (#25), Llama 4 Scout: 66.0 (#212)
| Benchmark | GPT-5.4 | Llama 4 Scout |
|---|---|---|
| LMArena Instruction Following | 1469 | 1248 |
| IFEval | — | 81.8% |
Long Context GPT-5.4 leads
GPT-5.4: 50.3 (#8), Llama 4 Scout: 27.5 (#289)
| Benchmark | GPT-5.4 | Llama 4 Scout |
|---|---|---|
| LMArena Longer Query | 1474 | 1265 |
| Fiction.LiveBench | — | 36% |
| CL-bench | 27.9% | — |
| CL-bench Life | 21.7% | — |
Writing & Preference GPT-5.4 leads
GPT-5.4: 72.1 (#18), Llama 4 Scout: 37.1 (#256)
| Benchmark | GPT-5.4 | Llama 4 Scout |
|---|---|---|
| LMArena Text | 1469 | 1279 |
| LMArena Creative Writing | 1439 | 1249 |
| EQ-Bench Creative Writing | 1840 | 783 |
| LMArena Multi-Turn | 1482 | 1281 |
| WildBench | — | 78% |
| EQ-Bench 4 | 1272 | — |
Frequently asked questions
Is GPT-5.4 better than Llama 4 Scout?
GPT-5.4 is the stronger model overall, scoring 59.5 to 27.4 on the Noometry Index. Llama 4 Scout costs 37× less per token, which makes it the better buy when GPT-5.4's lead doesn't matter for your workload.
Which is cheaper, GPT-5.4 or Llama 4 Scout?
Llama 4 Scout is cheaper. It lists at $0.10 per million input tokens and $0.30 per million output tokens; GPT-5.4 lists at $2.50 and $15.
Is GPT-5.4 or Llama 4 Scout better for coding?
GPT-5.4 scores higher on coding benchmarks: 52.8 versus 19.9 in the Noometry coding category.
Which has the bigger context window?
GPT-5.4 does, with 1.05M tokens against 128K.
How many benchmarks do GPT-5.4 and Llama 4 Scout share?
32 benchmarks have published results for both models. GPT-5.4 has 68 scored results on Noometry and Llama 4 Scout has 43.