Model comparison
Gemini 2.5 Flash vs GPT-5.4 mini
GPT-5.4 mini is the stronger model overall, scoring 45.0 to 39.3 on the Noometry Index. Gemini 2.5 Flash costs 2.0× less per token, which makes it the better buy when GPT-5.4 mini's lead doesn't matter for your workload.
Last verified . 33 shared benchmarks.
GoogleSummary
- They share 33 benchmarks with published results for both. Gemini 2.5 Flash scores higher in 5 categories and GPT-5.4 mini in 5 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where GPT-5.4 mini leads 51.5 to 36.4.
- The biggest single-benchmark swing is ARC-AGI-1: 33.3% for Gemini 2.5 Flash and 63.7% for GPT-5.4 mini.
- Gemini 2.5 Flash is cheaper at $0.30 / $2.50 per million input/output tokens, against $0.75 / $4.50 for GPT-5.4 mini.
- Gemini 2.5 Flash accepts more context: 1.05M tokens versus 400K.
Side by side
| Gemini 2.5 Flash | GPT-5.4 mini | |
|---|---|---|
| Provider | OpenAI | |
| Noometry Index | 39.3 | 45.0 |
| Released | 2025-04-17 | 2026-03-17 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 400K |
| Max output | 66K | 128K |
| Input $ / M tokens | $0.30 | $0.75 |
| Output $ / M tokens | $2.50 | $4.50 |
| Results tracked | 54 | 46 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.4 mini leads
Gemini 2.5 Flash: 35.8 (#220), GPT-5.4 mini: 45.2 (#72)
| Benchmark | Gemini 2.5 Flash | GPT-5.4 mini |
|---|---|---|
| WeirdML | 41.9% | 60.3% |
| LMArena Coding | 1424 | 1438 |
| ALE-Bench | 661.88 | 1,189 |
| FrontierCode | — | 27% |
| SWE-bench Verified (bash only) | 28.7% | — |
| Aider Polyglot | 55.1% | — |
| LMArena WebDev | — | 1397 |
| SciCode | — | 49.9% |
Agentic & Tool Use Too close to call
Gemini 2.5 Flash: 30.8 (#74), GPT-5.4 mini: 29.9 (#81)
| Benchmark | Gemini 2.5 Flash | GPT-5.4 mini |
|---|---|---|
| Terminal-Bench | 17.1% | — |
| Berkeley Function Calling Leaderboard | 56.2% | — |
| TheAgentCompany | 41.1% | — |
| DeepResearch Bench | — | 36.3% |
| BALROG | 33.5% | — |
| Vending-Bench 2 | 548.84 | — |
Reasoning GPT-5.4 mini leads
Gemini 2.5 Flash: 18.1 (#286), GPT-5.4 mini: 30.4 (#85)
| Benchmark | Gemini 2.5 Flash | GPT-5.4 mini |
|---|---|---|
| ARC-AGI-2 | 2.5% | 18.9% |
| Kagi LLM Benchmark | 56.8% | 37.9% |
| ARC-AGI-1 | 33.3% | 63.7% |
| CritPt | 1.1% | 10% |
| LMArena Hard Prompts | 1422 | 1424 |
| DTBench | 76.5% | 80% |
| LMCA | 27.5% | 40.8% |
| Epoch Capabilities Index | 143.03 | 148.84 |
| ForecastBench | 60.6 | 57 |
| SimpleBench | 41.2% | — |
| NYT Connections (extended) | — | 61.8% |
| Chess Puzzles | — | 24% |
| EnigmaEval | 2.7% | — |
| Thematic Generalization | — | 61.7% |
| Mystery Game Puzzles | — | 11% |
Math GPT-5.4 mini leads
Gemini 2.5 Flash: 39.9 (#98), GPT-5.4 mini: 45.5 (#75)
| Benchmark | Gemini 2.5 Flash | GPT-5.4 mini |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 73.1% | 88.9% |
| LMArena Math | 1415 | 1419 |
| FrontierMath (Feb 2025 set) | 4.8% | 28.3% |
| FrontierMath Tier 4 (v1) | 4.2% | 2.1% |
| FrontierMath (Tiers 1-3) | — | 51.2% |
| FrontierMath Tier 4 | — | 9.8% |
| ProofBench | — | 21% |
| Omni-MATH | 38.5% | — |
Knowledge GPT-5.4 mini leads
Gemini 2.5 Flash: 36.4 (#168), GPT-5.4 mini: 51.5 (#67)
| Benchmark | Gemini 2.5 Flash | GPT-5.4 mini |
|---|---|---|
| Vectara Hallucination Rate | 7.8% | 5.5% |
| LMArena Expert | 1426 | 1435 |
| GPQA Diamond | — | 86.9% |
| Humanity's Last Exam | 12.1% | — |
| SimpleQA Verified | — | 29.4% |
| MMLU-Pro | 63.9% | — |
| Confabulations | 16.8% | — |
| GPQA (HELM) | 39% | — |
Multimodal Gemini 2.5 Flash leads
Gemini 2.5 Flash: 41.8 (#32), GPT-5.4 mini: 39.7 (#56)
| Benchmark | Gemini 2.5 Flash | GPT-5.4 mini |
|---|---|---|
| LMArena Vision | 1253 | 1245 |
| GeoBench | 76% | — |
| VPCT | 46.2% | — |
| SpatialViz-Bench | 36.9% | — |
Multilingual Too close to call
Gemini 2.5 Flash: 52.3 (#88), GPT-5.4 mini: 51.9 (#96)
| Benchmark | Gemini 2.5 Flash | GPT-5.4 mini |
|---|---|---|
| LMArena Non-English | 1409 | 1405 |
| LMArena Chinese | 1450 | 1446 |
| LMArena French | 1433 | 1440 |
| LMArena German | 1418 | 1409 |
| LMArena Japanese | 1405 | 1374 |
| LMArena Korean | 1385 | 1368 |
| LMArena Russian | 1415 | 1417 |
| LMArena Spanish | 1421 | 1405 |
Instruction Following Gemini 2.5 Flash leads
Gemini 2.5 Flash: 75.7 (#54), GPT-5.4 mini: 74.1 (#102)
| Benchmark | Gemini 2.5 Flash | GPT-5.4 mini |
|---|---|---|
| LMArena Instruction Following | 1405 | 1405 |
| IFEval | 89.8% | — |
Long Context Gemini 2.5 Flash leads
Gemini 2.5 Flash: 47.5 (#17), GPT-5.4 mini: 43.0 (#112)
| Benchmark | Gemini 2.5 Flash | GPT-5.4 mini |
|---|---|---|
| LMArena Longer Query | 1419 | 1407 |
| Fiction.LiveBench | 77.8% | — |
Writing & Preference GPT-5.4 mini leads
Gemini 2.5 Flash: 53.8 (#157), GPT-5.4 mini: 64.0 (#58)
| Benchmark | Gemini 2.5 Flash | GPT-5.4 mini |
|---|---|---|
| LMArena Text | 1417 | 1412 |
| LMArena Creative Writing | 1400 | 1370 |
| EQ-Bench Creative Writing | 1137 | 1665 |
| LMArena Multi-Turn | 1408 | 1429 |
| Short-Story Creative Writing | 76.5% | — |
| WildBench | 81.7% | — |
Frequently asked questions
Is Gemini 2.5 Flash better than GPT-5.4 mini?
GPT-5.4 mini is the stronger model overall, scoring 45.0 to 39.3 on the Noometry Index. Gemini 2.5 Flash costs 2.0× less per token, which makes it the better buy when GPT-5.4 mini's lead doesn't matter for your workload.
Which is cheaper, Gemini 2.5 Flash or GPT-5.4 mini?
Gemini 2.5 Flash is cheaper. It lists at $0.30 per million input tokens and $2.50 per million output tokens; GPT-5.4 mini lists at $0.75 and $4.50.
Is Gemini 2.5 Flash or GPT-5.4 mini better for coding?
GPT-5.4 mini scores higher on coding benchmarks: 45.2 versus 35.8 in the Noometry coding category.
Which has the bigger context window?
Gemini 2.5 Flash does, with 1.05M tokens against 400K.
How many benchmarks do Gemini 2.5 Flash and GPT-5.4 mini share?
33 benchmarks have published results for both models. Gemini 2.5 Flash has 54 scored results on Noometry and GPT-5.4 mini has 46.