In short
How they differ.
Worked out from the figures below. Prices and specs are the makers' own, read Oct 20, 2026.
- Gemini 3.8 Flash is cheaper per output token: $3.75 against $15 per million tokens, so Kimi K3 costs 4 times as much.
- Kimi K3 can be downloaded and run on your own machine; Gemini 3.8 Flash can't.
- On the 8 independent boards that rank both, Gemini 3.8 Flash places higher on 6 and Kimi K3 on 2.
Side by side
The specs.
| Gemini 3.8 Flash | Kimi K3 | |
|---|---|---|
| Made by | Google · US | Moonshot AI · China |
| Released | Sep 2, 2026 | Not stated |
| Price per 1M tokens | $0.75 in / $3.75 out (introductory price) | $3 in / $15 out |
| Context window | 1M | 1M |
| Takes in | text, image, video, audio, pdf | text, image, video |
| Download and run it yourself | No | Yes (Kimi K3 License) |
| Size | Not published | 2.8T parameters, 104B active |
| Maker says | “Gemini 3.8 Flash is our most intelligent Flash model, engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows—all with the speed and cost efficiency of Flash.” | “Kimi K3 is an open-weight, native multimodal agentic model and our most capable model to date.” |
| Sources | Prices, Oct 20, 2026 · Description | Prices, Oct 20, 2026 · Description |
Independent tests
Where they place.
Position on each board we follow (1 is best). Epoch AI's and Vectara's scores are shown; the other boards are linked for their own figures.
| Board | Gemini 3.8 Flash | Kimi K3 |
|---|---|---|
| Epoch Capabilities IndexEpoch AI, Oct 20, 2026 | #19157.2 | #17157.42 |
| LMArena textLMArena, Oct 15, 2026 | #9 | #18 |
| LMArena creative writingLMArena, Oct 15, 2026 | #7 | Not listed |
| LMArena coding chatsLMArena, Oct 15, 2026 | Not listed | #12 |
| LMArena WebDevLMArena, Oct 19, 2026 | Not listed | #16 |
| LMArena mathLMArena, Oct 15, 2026 | #4 | #16 |
| LMArena visionLMArena, Oct 19, 2026 | #12 | Not listed |
| LMArena Agent ArenaLMArena, Oct 19, 2026 | #15 | Not listed |
| Humanity's Last ExamScale AI | #8 | #14 |
| SWE-Bench ProScale AI, Sep 22, 2026 | #12 | #4 |
| Remote Labor IndexScale AI | #7 | Not listed |
| Vals IndexVals AI, Oct 19, 2026 | #13 | #18 |
| Vibe Code BenchVals AI, Oct 18, 2026 | Not listed | #16 |
| ProofBenchVals AI, Oct 12, 2026 | Not listed | #12 |
| GPQA Diamond (run by Vals AI, archived)Vals AI, Sep 1, 2026 | #4 | #12 |
| MMMU Pro (run by Vals AI, archived)Vals AI, Sep 1, 2026 | #4 | #11 |
| CUA-bench (games)Vals AI, Oct 13, 2026 | #9 | Not listed |
More comparisons