Model comparison
Gemini 3 Flash Preview vs GPT-5.4
GPT-5.4 is the stronger model overall, scoring 59.4 to 52.3 on the Noometry Index. Gemini 3 Flash Preview costs 5.0× less per token, which makes it the better buy when GPT-5.4's lead doesn't matter for your workload.
Last verified . 49 shared benchmarks.
GoogleSummary
- They share 49 benchmarks with published results for both. Gemini 3 Flash Preview scores higher in 1 category and GPT-5.4 in 9 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.4 leads 73.5 to 51.7.
- The biggest single-benchmark swing is ProofBench: 15% for Gemini 3 Flash Preview and 56% for GPT-5.4.
- Gemini 3 Flash Preview is cheaper at $0.50 / $3 per million input/output tokens, against $2.50 / $15 for GPT-5.4.
- GPT-5.4 accepts more context: 1.05M tokens versus 1.05M.
Side by side
| Gemini 3 Flash Preview | GPT-5.4 | |
|---|---|---|
| Provider | OpenAI | |
| Noometry Index | 52.3 | 59.4 |
| Released | 2025-12-17 | 2026-03-05 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 1.05M |
| Max output | 66K | 128K |
| Input $ / M tokens | $0.50 | $2.50 |
| Output $ / M tokens | $3 | $15 |
| Results tracked | 59 | 68 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.4 leads
Gemini 3 Flash Preview: 50.9 (#42), GPT-5.4: 52.6 (#33)
| Benchmark | Gemini 3 Flash Preview | GPT-5.4 |
|---|---|---|
| SWE-bench Verified | 75.4% | 76.9% |
| LMArena WebDev | 1439 | 1465 |
| GSO | 9.8% | 31.4% |
| WeirdML | 61.6% | 77.7% |
| LMArena Coding | 1460 | 1497 |
| ALE-Bench | 1,367 | 1,607 |
| DeepSWE | — | 51.8% |
| SWE-bench Verified (bash only) | 75.8% | — |
| SWE-bench Multilingual | 72.7% | — |
| SciCode | — | 56.6% |
| MirrorCode | — | 15.6% |
| AlgoTune | — | 1.85 |
Agentic & Tool Use GPT-5.4 leads
Gemini 3 Flash Preview: 38.7 (#29), GPT-5.4: 46.5 (#13)
| Benchmark | Gemini 3 Flash Preview | GPT-5.4 |
|---|---|---|
| Terminal-Bench | 64.3% | 81.8% |
| τ²-bench Banking | 27.3% | 39.4% |
| DeepResearch Bench | 49.8% | 35.1% |
| LMArena Search | 1198 | 1197 |
| Vending-Bench 2 | 3,635 | 6,144 |
| APEX-Agents | — | 52.4% |
| τ²-bench Airline | 82.5% | — |
| τ²-bench Retail | 76.8% | — |
| τ²-bench Telecom | 91.2% | — |
| PostTrainBench | — | 19% |
| BALROG | 48.1% | — |
| GBAEval | — | 45.1% |
| GDP.pdf | 10% | — |
| METR Time Horizons | — | 74.3% |
Reasoning GPT-5.4 leads
Gemini 3 Flash Preview: 49.2 (#37), GPT-5.4: 61.8 (#19)
| Benchmark | Gemini 3 Flash Preview | GPT-5.4 |
|---|---|---|
| ARC-AGI-2 | 33.6% | 74% |
| NYT Connections (extended) | 83.1% | 91.3% |
| ARC-AGI-1 | 84.7% | 93.7% |
| Chess Puzzles | 40% | 44% |
| LMArena Hard Prompts | 1465 | 1485 |
| Mystery Game Puzzles | 26% | 37% |
| DTBench | 89.1% | 94.4% |
| LMCA | 43.1% | 52% |
| Epoch Capabilities Index | 151.8 | 156.81 |
| ForecastBench | 58.5 | 59.5 |
| SimpleBench | 61.1% | — |
| Kagi LLM Benchmark | — | 63.8% |
| CritPt | — | 23.4% |
| EnigmaEval | — | 16% |
| Thematic Generalization | — | 80% |
| EBR-Bench | — | 25.4% |
Math GPT-5.4 leads
Gemini 3 Flash Preview: 51.7 (#55), GPT-5.4: 73.5 (#19)
| Benchmark | Gemini 3 Flash Preview | GPT-5.4 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 51.2% | 78.6% |
| FrontierMath Tier 4 | 17.1% | 49% |
| MathArena Final-Answer Competitions | 67.6% | 83.1% |
| OTIS Mock AIME 2024-2025 | 95.6% | 97.8% |
| ProofBench | 15% | 56% |
| LMArena Math | 1473 | 1488 |
| FrontierMath (Feb 2025 set) | 35.6% | 47.6% |
| FrontierMath Tier 4 (v1) | 4.2% | 27.1% |
Knowledge GPT-5.4 leads
Gemini 3 Flash Preview: 58.8 (#33), GPT-5.4: 65.3 (#14)
| Benchmark | Gemini 3 Flash Preview | GPT-5.4 |
|---|---|---|
| GPQA Diamond | 89.4% | 93.3% |
| SimpleQA Verified | 66.8% | 45.1% |
| Vectara Hallucination Rate | 13.5% | 7% |
| LMArena Expert | 1462 | 1507 |
| Humanity's Last Exam | — | 36.2% |
Multimodal Gemini 3 Flash Preview leads
Gemini 3 Flash Preview: 45.5 (#16), GPT-5.4: 43.7 (#20)
| Benchmark | Gemini 3 Flash Preview | GPT-5.4 |
|---|---|---|
| LMArena Vision | 1285 | 1303 |
| Blueprint-Bench 2 | 0% | 27.1% |
| LMArena Document | 1413 | 1471 |
| GeoBench | 88% | — |
| VPCT | 72.6% | — |
| Furniture Assembly | — | 37.5% |
Multilingual Too close to call
Gemini 3 Flash Preview: 55.7 (#27), GPT-5.4: 56.2 (#23)
| Benchmark | Gemini 3 Flash Preview | GPT-5.4 |
|---|---|---|
| LMArena Non-English | 1458 | 1465 |
| LMArena Chinese | 1511 | 1519 |
| LMArena French | 1477 | 1493 |
| LMArena German | 1497 | 1472 |
| LMArena Japanese | 1489 | 1485 |
| LMArena Korean | 1443 | 1448 |
| LMArena Russian | 1480 | 1480 |
| LMArena Spanish | 1469 | 1454 |
Instruction Following GPT-5.4 leads
Gemini 3 Flash Preview: 75.7 (#56), GPT-5.4: 77.1 (#27)
| Benchmark | Gemini 3 Flash Preview | GPT-5.4 |
|---|---|---|
| LMArena Instruction Following | 1437 | 1469 |
Long Context GPT-5.4 leads
Gemini 3 Flash Preview: 44.4 (#67), GPT-5.4: 50.3 (#8)
| Benchmark | Gemini 3 Flash Preview | GPT-5.4 |
|---|---|---|
| LMArena Longer Query | 1452 | 1473 |
| CL-bench | — | 27.9% |
| CL-bench Life | — | 21.7% |
Writing & Preference GPT-5.4 leads
Gemini 3 Flash Preview: 65.5 (#45), GPT-5.4: 71.9 (#17)
| Benchmark | Gemini 3 Flash Preview | GPT-5.4 |
|---|---|---|
| LMArena Text | 1466 | 1469 |
| LMArena Creative Writing | 1457 | 1439 |
| LMArena Multi-Turn | 1471 | 1482 |
| EQ-Bench Creative Writing | — | 1840 |
| EQ-Bench 4 | — | 1272 |
Frequently asked questions
Is Gemini 3 Flash Preview better than GPT-5.4?
GPT-5.4 is the stronger model overall, scoring 59.4 to 52.3 on the Noometry Index. Gemini 3 Flash Preview costs 5.0× less per token, which makes it the better buy when GPT-5.4's lead doesn't matter for your workload.
Which is cheaper, Gemini 3 Flash Preview or GPT-5.4?
Gemini 3 Flash Preview is cheaper. It lists at $0.50 per million input tokens and $3 per million output tokens; GPT-5.4 lists at $2.50 and $15.
Is Gemini 3 Flash Preview or GPT-5.4 better for coding?
GPT-5.4 scores higher on coding benchmarks: 52.6 versus 50.9 in the Noometry coding category.
Which has the bigger context window?
GPT-5.4 does, with 1.05M tokens against 1.05M.
How many benchmarks do Gemini 3 Flash Preview and GPT-5.4 share?
49 benchmarks have published results for both models. Gemini 3 Flash Preview has 59 scored results on Noometry and GPT-5.4 has 68.