Model comparison
Llama 4 Maverick vs Step 3.5 Flash
Step 3.5 Flash is the stronger model overall, scoring 42.3 to 30.9 on the Noometry Index.
Last verified . 18 shared benchmarks.
Meta
StepFunSummary
- They share 18 benchmarks with published results for both. Llama 4 Maverick scores higher in 0 categories and Step 3.5 Flash in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where Step 3.5 Flash leads 58.8 to 38.8.
- The biggest single-benchmark swing is NYT Connections (extended): 8% for Llama 4 Maverick and 28.4% for Step 3.5 Flash.
- Step 3.5 Flash is cheaper at $0.10 / $0.30 per million input/output tokens, against $0.19 / $0.65 for Llama 4 Maverick.
- Step 3.5 Flash accepts more context: 256K tokens versus 128K.
Side by side
| Llama 4 Maverick | Step 3.5 Flash | |
|---|---|---|
| Provider | Meta | StepFun |
| Noometry Index | 30.9 | 42.3 |
| Released | 2025-04-05 | 2026-01-29 |
| Weights | Open | Open |
| Context window | 128K | 256K |
| Max output | 4K | 256K |
| Input $ / M tokens | $0.19 | $0.10 |
| Output $ / M tokens | $0.65 | $0.30 |
| Results tracked | 54 | 19 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Step 3.5 Flash leads
Llama 4 Maverick: 26.6 (#324), Step 3.5 Flash: 42.4 (#105)
| Benchmark | Llama 4 Maverick | Step 3.5 Flash |
|---|---|---|
| LMArena Coding | 1302 | 1436 |
| SWE-bench Verified (bash only) | 21% | — |
| Aider Polyglot | 15.6% | — |
| SciCode | 33.1% | — |
| WeirdML | 24.5% | — |
| BigCodeBench Instruct | 49.7% | — |
| BigCodeBench Complete | 61.4% | — |
| ALE-Bench | 172.97 | — |
Agentic & Tool Use Not comparable
Llama 4 Maverick: 28.2 (#91), Step 3.5 Flash: —
| Benchmark | Llama 4 Maverick | Step 3.5 Flash |
|---|---|---|
| Berkeley Function Calling Leaderboard | 37.3% | — |
Reasoning Step 3.5 Flash leads
Llama 4 Maverick: 10.1 (#342), Step 3.5 Flash: 22.2 (#202)
| Benchmark | Llama 4 Maverick | Step 3.5 Flash |
|---|---|---|
| NYT Connections (extended) | 8% | 28.4% |
| LMArena Hard Prompts | 1281 | 1411 |
| ARC-AGI-2 | 0% | — |
| SimpleBench | 27.7% | — |
| Kagi LLM Benchmark | 55.9% | — |
| ARC-AGI-1 | 4.4% | — |
| CritPt | 0% | — |
| EnigmaEval | 0.6% | — |
| DTBench | 61.9% | — |
| LMCA | 15.9% | — |
| Epoch Capabilities Index | 132.2 | — |
| ForecastBench | 57.5 | — |
Math Step 3.5 Flash leads
Llama 4 Maverick: 26.0 (#262), Step 3.5 Flash: 42.6 (#84)
| Benchmark | Llama 4 Maverick | Step 3.5 Flash |
|---|---|---|
| LMArena Math | 1299 | 1408 |
| MathArena Final-Answer Competitions | — | 66.8% |
| OTIS Mock AIME 2024-2025 | 20.6% | — |
| Omni-MATH | 42.2% | — |
| MATH Level 5 | 73% | — |
| FrontierMath (Feb 2025 set) | 0.7% | — |
Knowledge Step 3.5 Flash leads
Llama 4 Maverick: 33.4 (#204), Step 3.5 Flash: 39.6 (#132)
| Benchmark | Llama 4 Maverick | Step 3.5 Flash |
|---|---|---|
| LMArena Expert | 1259 | 1421 |
| GPQA Diamond | 67% | — |
| Humanity's Last Exam | 5.7% | — |
| MMLU-Pro | 81% | — |
| Confabulations | 22.6% | — |
| Vectara Hallucination Rate | 8.2% | — |
| GPQA (HELM) | 65% | — |
Multimodal Not comparable
Llama 4 Maverick: 31.6 (#105), Step 3.5 Flash: —
| Benchmark | Llama 4 Maverick | Step 3.5 Flash |
|---|---|---|
| LMArena Vision | 1142 | — |
| GeoBench | 52% | — |
| SpatialViz-Bench | 31.8% | — |
Multilingual Step 3.5 Flash leads
Llama 4 Maverick: 42.2 (#195), Step 3.5 Flash: 50.5 (#119)
| Benchmark | Llama 4 Maverick | Step 3.5 Flash |
|---|---|---|
| LMArena Non-English | 1269 | 1385 |
| LMArena Chinese | 1277 | 1447 |
| LMArena French | 1259 | 1421 |
| LMArena German | 1291 | 1405 |
| LMArena Japanese | 1207 | 1354 |
| LMArena Korean | 1203 | 1352 |
| LMArena Russian | 1286 | 1385 |
| LMArena Spanish | 1293 | 1419 |
Instruction Following Step 3.5 Flash leads
Llama 4 Maverick: 71.7 (#146), Step 3.5 Flash: 73.1 (#124)
| Benchmark | Llama 4 Maverick | Step 3.5 Flash |
|---|---|---|
| LMArena Instruction Following | 1267 | 1385 |
| IFEval | 90.8% | — |
Long Context Step 3.5 Flash leads
Llama 4 Maverick: 31.4 (#279), Step 3.5 Flash: 42.8 (#117)
| Benchmark | Llama 4 Maverick | Step 3.5 Flash |
|---|---|---|
| LMArena Longer Query | 1280 | 1402 |
| Fiction.LiveBench | 46.2% | — |
Writing & Preference Step 3.5 Flash leads
Llama 4 Maverick: 38.8 (#252), Step 3.5 Flash: 58.8 (#113)
| Benchmark | Llama 4 Maverick | Step 3.5 Flash |
|---|---|---|
| LMArena Text | 1287 | 1403 |
| LMArena Creative Writing | 1267 | 1357 |
| LMArena Multi-Turn | 1289 | 1405 |
| Short-Story Creative Writing | 62% | — |
| EQ-Bench Creative Writing | 860 | — |
| WildBench | 80% | — |
Frequently asked questions
Is Llama 4 Maverick better than Step 3.5 Flash?
Step 3.5 Flash is the stronger model overall, scoring 42.3 to 30.9 on the Noometry Index.
Which is cheaper, Llama 4 Maverick or Step 3.5 Flash?
Step 3.5 Flash is cheaper. It lists at $0.10 per million input tokens and $0.30 per million output tokens; Llama 4 Maverick lists at $0.19 and $0.65.
Is Llama 4 Maverick or Step 3.5 Flash better for coding?
Step 3.5 Flash scores higher on coding benchmarks: 42.4 versus 26.6 in the Noometry coding category.
Which has the bigger context window?
Step 3.5 Flash does, with 256K tokens against 128K.
How many benchmarks do Llama 4 Maverick and Step 3.5 Flash share?
18 benchmarks have published results for both models. Llama 4 Maverick has 54 scored results on Noometry and Step 3.5 Flash has 19.