Model comparison
Gemini 3.8 Flash vs Gemini 3 Flash Preview
Gemini 3.8 Flash is the stronger model overall, scoring 61.8 to 52.3 on the Noometry Index.
Last verified . 38 shared benchmarks.
GoogleSummary
- They share 38 benchmarks with published results for both. Gemini 3.8 Flash scores higher in 9 categories and Gemini 3 Flash Preview in 1 category; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Gemini 3.8 Flash leads 76.9 to 49.2.
- The biggest single-benchmark swing is ARC-AGI-2: 89.2% for Gemini 3.8 Flash and 33.6% for Gemini 3 Flash Preview.
- Gemini 3 Flash Preview is cheaper at $0.50 / $3 per million input/output tokens, against $0.75 / $3.75 for Gemini 3.8 Flash.
Side by side
| Gemini 3.8 Flash | Gemini 3 Flash Preview | |
|---|---|---|
| Provider | ||
| Noometry Index | 61.8 | 52.3 |
| Released | 2026-09-02 | 2025-12-17 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 1.05M |
| Max output | 66K | 66K |
| Input $ / M tokens | $0.75 | $0.50 |
| Output $ / M tokens | $3.75 | $3 |
| Results tracked | 50 | 59 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Gemini 3.8 Flash leads
Gemini 3.8 Flash: 59.2 (#15), Gemini 3 Flash Preview: 50.9 (#42)
| Benchmark | Gemini 3.8 Flash | Gemini 3 Flash Preview |
|---|---|---|
| LMArena WebDev | 1584 | 1439 |
| WeirdML | 84.8% | 61.6% |
| LMArena Coding | 1510 | 1460 |
| ALE-Bench | 1,270 | 1,367 |
| SWE-bench Verified | — | 75.4% |
| DeepSWE | 73.8% | — |
| FrontierCode | 41.2% | — |
| SWE-bench Verified (bash only) | — | 75.8% |
| CursorBench | 39.6% | — |
| SWE-bench Multilingual | — | 72.7% |
| FrontierSWE | 19.6% | — |
| SciCode | 56.6% | — |
| GSO | — | 9.8% |
Agentic & Tool Use Gemini 3.8 Flash leads
Gemini 3.8 Flash: 41.8 (#21), Gemini 3 Flash Preview: 38.7 (#29)
| Benchmark | Gemini 3.8 Flash | Gemini 3 Flash Preview |
|---|---|---|
| GDP.pdf | 23.4% | 10% |
| Vending-Bench 2 | 5,094 | 3,635 |
| Terminal-Bench | — | 64.3% |
| APEX-Agents | 64.3% | — |
| Remote Labor Index | 5.8% | — |
| τ²-bench Airline | — | 82.5% |
| τ²-bench Banking | — | 27.3% |
| τ²-bench Retail | — | 76.8% |
| τ²-bench Telecom | — | 91.2% |
| DeepResearch Bench | — | 49.8% |
| BALROG | — | 48.1% |
| LMArena Search | — | 1198 |
Reasoning Gemini 3.8 Flash leads
Gemini 3.8 Flash: 76.9 (#5), Gemini 3 Flash Preview: 49.2 (#37)
| Benchmark | Gemini 3.8 Flash | Gemini 3 Flash Preview |
|---|---|---|
| ARC-AGI-2 | 89.2% | 33.6% |
| NYT Connections (extended) | 97.4% | 83.1% |
| ARC-AGI-1 | 98.5% | 84.7% |
| Chess Puzzles | 61% | 40% |
| LMArena Hard Prompts | 1508 | 1465 |
| Mystery Game Puzzles | 47% | 26% |
| DTBench | 95.7% | 89.1% |
| LMCA | 52.9% | 43.1% |
| Epoch Capabilities Index | 156.71 | 151.8 |
| SimpleBench | — | 61.1% |
| CritPt | 18.3% | — |
| Surface Evolver Bench | 76.9% | — |
| ForecastBench | — | 58.5 |
Math Gemini 3.8 Flash leads
Gemini 3.8 Flash: 65.3 (#28), Gemini 3 Flash Preview: 51.7 (#55)
| Benchmark | Gemini 3.8 Flash | Gemini 3 Flash Preview |
|---|---|---|
| FrontierMath (Tiers 1-3) | 68.4% | 51.2% |
| FrontierMath Tier 4 | 22% | 17.1% |
| OTIS Mock AIME 2024-2025 | 98.9% | 95.6% |
| ProofBench | 48% | 15% |
| LMArena Math | 1528 | 1473 |
| MathArena Final-Answer Competitions | — | 67.6% |
| FrontierMath (Feb 2025 set) | — | 35.6% |
| FrontierMath Tier 4 (v1) | — | 4.2% |
Knowledge Gemini 3.8 Flash leads
Gemini 3.8 Flash: 74.8 (#2), Gemini 3 Flash Preview: 58.8 (#33)
| Benchmark | Gemini 3.8 Flash | Gemini 3 Flash Preview |
|---|---|---|
| GPQA Diamond | 95.4% | 89.4% |
| SimpleQA Verified | 69.7% | 66.8% |
| LMArena Expert | 1524 | 1462 |
| Humanity's Last Exam | 44.5% | — |
| Vectara Hallucination Rate | — | 13.5% |
Multimodal Gemini 3 Flash Preview leads
Gemini 3.8 Flash: 40.7 (#45), Gemini 3 Flash Preview: 45.5 (#16)
| Benchmark | Gemini 3.8 Flash | Gemini 3 Flash Preview |
|---|---|---|
| LMArena Vision | 1314 | 1285 |
| Blueprint-Bench 2 | 38.6% | 0% |
| GeoBench | — | 88% |
| VPCT | — | 72.6% |
| Furniture Assembly | 31.7% | — |
| LMArena Document | — | 1413 |
Multilingual Gemini 3.8 Flash leads
Gemini 3.8 Flash: 58.0 (#5), Gemini 3 Flash Preview: 55.7 (#27)
| Benchmark | Gemini 3.8 Flash | Gemini 3 Flash Preview |
|---|---|---|
| LMArena Non-English | 1491 | 1458 |
| LMArena Chinese | 1554 | 1511 |
| LMArena French | 1498 | 1477 |
| LMArena German | 1493 | 1497 |
| LMArena Japanese | 1502 | 1489 |
| LMArena Korean | 1459 | 1443 |
| LMArena Russian | 1515 | 1480 |
| LMArena Spanish | 1485 | 1469 |
Instruction Following Gemini 3.8 Flash leads
Gemini 3.8 Flash: 78.0 (#13), Gemini 3 Flash Preview: 75.7 (#56)
| Benchmark | Gemini 3.8 Flash | Gemini 3 Flash Preview |
|---|---|---|
| LMArena Instruction Following | 1490 | 1437 |
Long Context Gemini 3.8 Flash leads
Gemini 3.8 Flash: 46.3 (#24), Gemini 3 Flash Preview: 44.4 (#67)
| Benchmark | Gemini 3.8 Flash | Gemini 3 Flash Preview |
|---|---|---|
| LMArena Longer Query | 1508 | 1452 |
Writing & Preference Gemini 3.8 Flash leads
Gemini 3.8 Flash: 72.2 (#15), Gemini 3 Flash Preview: 65.5 (#45)
| Benchmark | Gemini 3.8 Flash | Gemini 3 Flash Preview |
|---|---|---|
| LMArena Text | 1499 | 1466 |
| LMArena Creative Writing | 1492 | 1457 |
| LMArena Multi-Turn | 1501 | 1471 |
| EQ-Bench Creative Writing | 1748 | — |
Frequently asked questions
Is Gemini 3.8 Flash better than Gemini 3 Flash Preview?
Gemini 3.8 Flash is the stronger model overall, scoring 61.8 to 52.3 on the Noometry Index.
Which is cheaper, Gemini 3.8 Flash or Gemini 3 Flash Preview?
Gemini 3 Flash Preview is cheaper. It lists at $0.50 per million input tokens and $3 per million output tokens; Gemini 3.8 Flash lists at $0.75 and $3.75.
Is Gemini 3.8 Flash or Gemini 3 Flash Preview better for coding?
Gemini 3.8 Flash scores higher on coding benchmarks: 59.2 versus 50.9 in the Noometry coding category.
Which has the bigger context window?
Both accept 1.05M tokens.
How many benchmarks do Gemini 3.8 Flash and Gemini 3 Flash Preview share?
38 benchmarks have published results for both models. Gemini 3.8 Flash has 50 scored results on Noometry and Gemini 3 Flash Preview has 59.