| <!DOCTYPE html> |
| <html lang="en"> |
| <head> |
| <meta charset="utf-8"> |
| <meta name="viewport" content="width=device-width, initial-scale=1"> |
| <title>compare · xai</title> |
| <link rel="stylesheet" href="style.css"> |
| </head> |
| <body> |
|
|
| <h1>Inkling vs Kimi K3</h1> |
| <p class="meta">primary sources only · not invented parity</p> |
| <nav class="meta"> |
| <a href="index.html">home</a> |
| <a href="inkling.html">Inkling</a> |
| <a href="kimi-k3.html">Kimi K3</a> |
| <a href="compare.html">compare</a> |
| <a href="study.html">study</a> |
| <a href="sources.html">sources</a> |
| <a href="msc.html">MSc</a> |
| </nav> |
|
|
| <hr> |
|
|
| <table> |
| <tr><th>axis</th><th>Inkling</th><th>Kimi K3</th></tr> |
| <tr><td>total</td><td>~975B</td><td>2.8T</td></tr> |
| <tr><td>active</td><td>~41B</td><td>undisclosed</td></tr> |
| <tr><td>MoE</td><td>6/256+2 sink</td><td>16/896 LatentMoE</td></tr> |
| <tr><td>ctx</td><td>1M</td><td>1M</td></tr> |
| <tr><td>length</td><td>relative + SWA/global</td><td>KDA</td></tr> |
| <tr><td>depth</td><td>SConv</td><td>AttnRes</td></tr> |
| <tr><td>mm</td><td>image+audio</td><td>vision+video</td></tr> |
| <tr><td>think</td><td>effort 0–0.99</td><td>always-on · max</td></tr> |
| <tr><td>code</td><td>day-0 transformers</td><td>weights ETA</td></tr> |
| <tr><td>risk</td><td>app-layer safety</td><td>preserved thinking</td></tr> |
| </table> |
|
|
| <p> |
| Prefer <a href="inkling.html">Inkling</a> for open weights + HF code. |
| Prefer <a href="kimi-k3.html">Kimi K3</a> for live 1M API + KDA theory. |
| </p> |
|
|
| <p class="meta">gaps: K3 active params · layer table · matched agentic benches</p> |
| <p class="meta"><a href="index.html">← home</a></p> |
| </body> |
| </html> |
| |
|
|
