Model comparison
Gemini 2.5 Flash-Lite vs GPT-5.4
GPT-5.4 is the stronger model overall, scoring 59.5 to 36.4 on the Noometry Index. Gemini 2.5 Flash-Lite costs 32× less per token, which makes it the better buy when GPT-5.4's lead doesn't matter for your workload.
Last verified . 25 shared benchmarks.
GoogleSummary
- They share 25 benchmarks with published results for both. Gemini 2.5 Flash-Lite scores higher in 0 categories and GPT-5.4 in 10 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.4 leads 61.7 to 21.5.
- The biggest single-benchmark swing is WeirdML: 35.2% for Gemini 2.5 Flash-Lite and 77.7% for GPT-5.4.
- Gemini 2.5 Flash-Lite is cheaper at $0.10 / $0.40 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 2.5 Flash-Lite | GPT-5.4 | |
|---|---|---|
| Provider | OpenAI | |
| Noometry Index | 36.4 | 59.5 |
| Released | 2025-06-17 | 2026-03-05 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 1.05M |
| Max output | 66K | 128K |
| Input $ / M tokens | $0.10 | $2.50 |
| Output $ / M tokens | $0.40 | $15 |
| Results tracked | 33 | 68 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.4 leads
Gemini 2.5 Flash-Lite: 38.0 (#167), GPT-5.4: 52.8 (#28)
| Benchmark | Gemini 2.5 Flash-Lite | GPT-5.4 |
|---|---|---|
| WeirdML | 35.2% | 77.7% |
| LMArena Coding | 1373 | 1497 |
| ALE-Bench | 325.9 | 1,607 |
| SWE-bench Verified | — | 76.9% |
| DeepSWE | — | 51.8% |
| LMArena WebDev | — | 1465 |
| SciCode | — | 56.6% |
| GSO | — | 31.4% |
| MirrorCode | — | 15.6% |
| AlgoTune | — | 1.85 |
Agentic & Tool Use GPT-5.4 leads
Gemini 2.5 Flash-Lite: 27.8 (#89), GPT-5.4: 46.5 (#12)
| Benchmark | Gemini 2.5 Flash-Lite | GPT-5.4 |
|---|---|---|
| Terminal-Bench | — | 81.8% |
| APEX-Agents | — | 52.4% |
| Berkeley Function Calling Leaderboard | 36.9% | — |
| 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
Gemini 2.5 Flash-Lite: 21.5 (#196), GPT-5.4: 61.7 (#17)
| Benchmark | Gemini 2.5 Flash-Lite | GPT-5.4 |
|---|---|---|
| Kagi LLM Benchmark | 40.5% | 63.8% |
| LMArena Hard Prompts | 1377 | 1485 |
| DTBench | 62.8% | 94.4% |
| LMCA | 18.1% | 52% |
| Epoch Capabilities Index | 133.93 | 156.92 |
| ARC-AGI-2 | — | 74% |
| NYT Connections (extended) | — | 91.3% |
| ARC-AGI-1 | — | 93.7% |
| CritPt | — | 23.4% |
| Chess Puzzles | — | 44% |
| EnigmaEval | — | 16% |
| Thematic Generalization | — | 80% |
| EBR-Bench | — | 25.4% |
| Mystery Game Puzzles | — | 37% |
| ForecastBench | — | 59.5 |
Math GPT-5.4 leads
Gemini 2.5 Flash-Lite: 35.5 (#144), GPT-5.4: 73.3 (#14)
| Benchmark | Gemini 2.5 Flash-Lite | GPT-5.4 |
|---|---|---|
| LMArena Math | 1373 | 1488 |
| FrontierMath (Tiers 1-3) | — | 78.6% |
| FrontierMath Tier 4 | — | 49% |
| MathArena Final-Answer Competitions | — | 83.1% |
| OTIS Mock AIME 2024-2025 | — | 97.8% |
| ProofBench | — | 56% |
| Omni-MATH | 48% | — |
| FrontierMath (Feb 2025 set) | — | 47.6% |
| FrontierMath Tier 4 (v1) | — | 27.1% |
Knowledge GPT-5.4 leads
Gemini 2.5 Flash-Lite: 32.1 (#206), GPT-5.4: 65.3 (#13)
| Benchmark | Gemini 2.5 Flash-Lite | GPT-5.4 |
|---|---|---|
| Vectara Hallucination Rate | 3.3% | 7% |
| LMArena Expert | 1373 | 1507 |
| GPQA Diamond | — | 93.3% |
| Humanity's Last Exam | — | 36.2% |
| SimpleQA Verified | — | 45.1% |
| MMLU-Pro | 53.7% | — |
| GPQA (HELM) | 30.9% | — |
Multimodal GPT-5.4 leads
Gemini 2.5 Flash-Lite: 28.6 (#109), GPT-5.4: 43.8 (#19)
| Benchmark | Gemini 2.5 Flash-Lite | GPT-5.4 |
|---|---|---|
| LMArena Vision | 1198 | 1302 |
| VPCT | 30% | — |
| Blueprint-Bench 2 | — | 27.1% |
| Furniture Assembly | — | 37.5% |
| LMArena Document | — | 1471 |
Multilingual GPT-5.4 leads
Gemini 2.5 Flash-Lite: 49.5 (#129), GPT-5.4: 56.4 (#22)
| Benchmark | Gemini 2.5 Flash-Lite | GPT-5.4 |
|---|---|---|
| LMArena Non-English | 1369 | 1465 |
| LMArena Chinese | 1404 | 1520 |
| LMArena French | 1388 | 1493 |
| LMArena German | 1390 | 1472 |
| LMArena Japanese | 1359 | 1484 |
| LMArena Korean | 1358 | 1449 |
| LMArena Russian | 1373 | 1480 |
| LMArena Spanish | 1396 | 1453 |
Instruction Following GPT-5.4 leads
Gemini 2.5 Flash-Lite: 70.2 (#163), GPT-5.4: 77.3 (#25)
| Benchmark | Gemini 2.5 Flash-Lite | GPT-5.4 |
|---|---|---|
| LMArena Instruction Following | 1366 | 1469 |
| IFEval | 81% | — |
Long Context GPT-5.4 leads
Gemini 2.5 Flash-Lite: 33.3 (#257), GPT-5.4: 50.3 (#8)
| Benchmark | Gemini 2.5 Flash-Lite | GPT-5.4 |
|---|---|---|
| LMArena Longer Query | 1373 | 1474 |
| Fiction.LiveBench | 47.2% | — |
| CL-bench | — | 27.9% |
| CL-bench Life | — | 21.7% |
Writing & Preference GPT-5.4 leads
Gemini 2.5 Flash-Lite: 57.0 (#130), GPT-5.4: 72.1 (#18)
| Benchmark | Gemini 2.5 Flash-Lite | GPT-5.4 |
|---|---|---|
| LMArena Text | 1379 | 1469 |
| LMArena Creative Writing | 1366 | 1439 |
| LMArena Multi-Turn | 1367 | 1482 |
| EQ-Bench Creative Writing | — | 1840 |
| WildBench | 81.8% | — |
| EQ-Bench 4 | — | 1272 |
Frequently asked questions
Is Gemini 2.5 Flash-Lite better than GPT-5.4?
GPT-5.4 is the stronger model overall, scoring 59.5 to 36.4 on the Noometry Index. Gemini 2.5 Flash-Lite costs 32× 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 2.5 Flash-Lite or GPT-5.4?
Gemini 2.5 Flash-Lite is cheaper. It lists at $0.10 per million input tokens and $0.40 per million output tokens; GPT-5.4 lists at $2.50 and $15.
Is Gemini 2.5 Flash-Lite or GPT-5.4 better for coding?
GPT-5.4 scores higher on coding benchmarks: 52.8 versus 38.0 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 2.5 Flash-Lite and GPT-5.4 share?
25 benchmarks have published results for both models. Gemini 2.5 Flash-Lite has 33 scored results on Noometry and GPT-5.4 has 68.