Model comparison
Gemini 3.8 Flash vs Kimi K2.5
Gemini 3.8 Flash is the stronger model overall, scoring 61.8 to 48.1 on the Noometry Index. Kimi K2.5 costs 1.7× less per token, which makes it the better buy when Gemini 3.8 Flash's lead doesn't matter for your workload.
Last verified . 34 shared benchmarks.
GoogleSummary
- They share 34 benchmarks with published results for both. Gemini 3.8 Flash scores higher in 8 categories and Kimi K2.5 in 2 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Gemini 3.8 Flash leads 76.9 to 31.2.
- The biggest single-benchmark swing is ARC-AGI-2: 89.2% for Gemini 3.8 Flash and 11.8% for Kimi K2.5.
- Kimi K2.5 is cheaper at $0.45 / $2.25 per million input/output tokens, against $0.75 / $3.75 for Gemini 3.8 Flash.
- Gemini 3.8 Flash accepts more context: 1.05M tokens versus 262K.
- Kimi K2.5 has downloadable open weights; the other is API-only.
Side by side
| Gemini 3.8 Flash | Kimi K2.5 | |
|---|---|---|
| Provider | Moonshot AI | |
| Noometry Index | 61.8 | 48.1 |
| Released | 2026-09-02 | 2026-01-27 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 262K |
| Max output | 66K | 262K |
| Input $ / M tokens | $0.75 | $0.45 |
| Output $ / M tokens | $3.75 | $2.25 |
| Results tracked | 50 | 51 |
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), Kimi K2.5: 48.8 (#53)
| Benchmark | Gemini 3.8 Flash | Kimi K2.5 |
|---|---|---|
| LMArena WebDev | 1584 | 1437 |
| SciCode | 56.6% | 49% |
| WeirdML | 84.8% | 45.6% |
| LMArena Coding | 1510 | 1474 |
| ALE-Bench | 1,270 | 821.65 |
| SWE-bench Verified | — | 73.8% |
| DeepSWE | 73.8% | — |
| FrontierCode | 41.2% | — |
| SWE-bench Verified (bash only) | — | 70.8% |
| CursorBench | 39.6% | — |
| SWE-bench Multilingual | — | 67.3% |
| FrontierSWE | 19.6% | — |
Agentic & Tool Use Gemini 3.8 Flash leads
Gemini 3.8 Flash: 41.8 (#21), Kimi K2.5: 34.2 (#48)
| Benchmark | Gemini 3.8 Flash | Kimi K2.5 |
|---|---|---|
| Vending-Bench 2 | 5,094 | 1,198 |
| Terminal-Bench | — | 43.2% |
| APEX-Agents | 64.3% | — |
| Remote Labor Index | 5.8% | — |
| OSWorld | — | 63.3% |
| GDP.pdf | 23.4% | — |
Reasoning Gemini 3.8 Flash leads
Gemini 3.8 Flash: 76.9 (#5), Kimi K2.5: 31.2 (#80)
| Benchmark | Gemini 3.8 Flash | Kimi K2.5 |
|---|---|---|
| ARC-AGI-2 | 89.2% | 11.8% |
| NYT Connections (extended) | 97.4% | 69.9% |
| ARC-AGI-1 | 98.5% | 65.3% |
| CritPt | 18.3% | 3.1% |
| Chess Puzzles | 61% | 12% |
| LMArena Hard Prompts | 1508 | 1453 |
| Epoch Capabilities Index | 156.71 | 148.03 |
| SimpleBench | — | 46.8% |
| Kagi LLM Benchmark | — | 78.5% |
| EnigmaEval | — | 3.4% |
| Thematic Generalization | — | 69.4% |
| Mystery Game Puzzles | 47% | — |
| DTBench | 95.7% | — |
| LMCA | 52.9% | — |
| Surface Evolver Bench | 76.9% | — |
Math Gemini 3.8 Flash leads
Gemini 3.8 Flash: 65.3 (#28), Kimi K2.5: 51.8 (#53)
| Benchmark | Gemini 3.8 Flash | Kimi K2.5 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 98.9% | 92.2% |
| LMArena Math | 1528 | 1470 |
| FrontierMath (Tiers 1-3) | 68.4% | — |
| FrontierMath Tier 4 | 22% | — |
| MathArena Final-Answer Competitions | — | 62.3% |
| ProofBench | 48% | — |
| FrontierMath (Feb 2025 set) | — | 27.9% |
| FrontierMath Tier 4 (v1) | — | 4.2% |
Knowledge Gemini 3.8 Flash leads
Gemini 3.8 Flash: 74.8 (#2), Kimi K2.5: 53.6 (#56)
| Benchmark | Gemini 3.8 Flash | Kimi K2.5 |
|---|---|---|
| GPQA Diamond | 95.4% | 87.6% |
| Humanity's Last Exam | 44.5% | 24.4% |
| SimpleQA Verified | 69.7% | 34.3% |
| LMArena Expert | 1524 | 1466 |
| Vectara Hallucination Rate | — | 14.2% |
Multimodal Too close to call
Gemini 3.8 Flash: 40.7 (#45), Kimi K2.5: 41.1 (#39)
| Benchmark | Gemini 3.8 Flash | Kimi K2.5 |
|---|---|---|
| LMArena Vision | 1314 | 1269 |
| Blueprint-Bench 2 | 38.6% | — |
| Furniture Assembly | 31.7% | — |
| LMArena Document | — | 1430 |
Multilingual Gemini 3.8 Flash leads
Gemini 3.8 Flash: 58.0 (#5), Kimi K2.5: 53.9 (#53)
| Benchmark | Gemini 3.8 Flash | Kimi K2.5 |
|---|---|---|
| LMArena Non-English | 1491 | 1433 |
| LMArena Chinese | 1554 | 1495 |
| LMArena French | 1498 | 1454 |
| LMArena German | 1493 | 1441 |
| LMArena Japanese | 1502 | 1421 |
| LMArena Korean | 1459 | 1410 |
| LMArena Russian | 1515 | 1435 |
| LMArena Spanish | 1485 | 1450 |
Instruction Following Gemini 3.8 Flash leads
Gemini 3.8 Flash: 78.0 (#13), Kimi K2.5: 75.3 (#64)
| Benchmark | Gemini 3.8 Flash | Kimi K2.5 |
|---|---|---|
| LMArena Instruction Following | 1490 | 1431 |
Long Context Kimi K2.5 leads
Gemini 3.8 Flash: 46.3 (#24), Kimi K2.5: 52.1 (#7)
| Benchmark | Gemini 3.8 Flash | Kimi K2.5 |
|---|---|---|
| LMArena Longer Query | 1508 | 1445 |
| Fiction.LiveBench | — | 86.1% |
| CL-bench | — | 19.3% |
| CL-bench Life | — | 13.2% |
Writing & Preference Gemini 3.8 Flash leads
Gemini 3.8 Flash: 72.2 (#15), Kimi K2.5: 65.1 (#53)
| Benchmark | Gemini 3.8 Flash | Kimi K2.5 |
|---|---|---|
| LMArena Text | 1499 | 1445 |
| LMArena Creative Writing | 1492 | 1423 |
| EQ-Bench Creative Writing | 1748 | 1579 |
| LMArena Multi-Turn | 1501 | 1444 |
Frequently asked questions
Is Gemini 3.8 Flash better than Kimi K2.5?
Gemini 3.8 Flash is the stronger model overall, scoring 61.8 to 48.1 on the Noometry Index. Kimi K2.5 costs 1.7× less per token, which makes it the better buy when Gemini 3.8 Flash's lead doesn't matter for your workload.
Which is cheaper, Gemini 3.8 Flash or Kimi K2.5?
Kimi K2.5 is cheaper. It lists at $0.45 per million input tokens and $2.25 per million output tokens; Gemini 3.8 Flash lists at $0.75 and $3.75.
Is Gemini 3.8 Flash or Kimi K2.5 better for coding?
Gemini 3.8 Flash scores higher on coding benchmarks: 59.2 versus 48.8 in the Noometry coding category.
Which has the bigger context window?
Gemini 3.8 Flash does, with 1.05M tokens against 262K.
How many benchmarks do Gemini 3.8 Flash and Kimi K2.5 share?
34 benchmarks have published results for both models. Gemini 3.8 Flash has 50 scored results on Noometry and Kimi K2.5 has 51.