External independent evaluation
Moonshot AI / 月之暗面 · Released 2026-07-16
Kimi K3
Native vision and long context support coding and knowledge work, with an SVG and image description making capabilities concrete; K3 and Code's K2.8 Preview are parallel paths.
Original model explanation ↗No comparable leaderboard score is included yet. Explore the evaluations and examples below.
Evaluations: conditions and results
External independent evaluation · separate from the intelligence index
Task profiles within a shared benchmark release ↗
Official calling guide · API defaults to max
Reasoning effort and long context in the general API ↗
Official serving routes · Code's k3 defaults to high
How Code's k3 and K2.8 Preview differ ↗
Official model card · self-hosting materials
Open weights and the model's current license ↗
How this capability connects to works and uses
Public creator run · one example from July 16
Generate an SVG, then return its image for a description ↗
Official integration example · tool and model roles
How local OpenCode connects to cloud Kimi →
In AWS’s September 23 example, Kimi K3 is asked to analyze two program tasks waiting on each other.
Moonshot AI / 月之暗面 model evolution
- Kimi K2.6
Long-horizon coding and agent work advance with open weights, providing a previous-generation reference for later general models.
Original release ↗ - Kimi K2.7 Code
The coding branch launches with open weights, exploring longer engineering work and reasoning-token trade-offs.
Original release ↗ - Kimi K3
Native vision and 1M context reach long-horizon coding and knowledge work; the full weights follow on July 27.
Original release ↗ - Kimi Code → K2.8 Preview
The standard kimi-for-coding endpoint moves to K2.8 Preview with more efficient reasoning, alongside the K3 flagship.
Original release ↗ - Kimi K3.1 Pro / Highspeed · Code上线
Code records rollout on October 23, followed by the October 24 research-index entry. Separate 1M eligibility, thinking and quotas for Pro/Highspeed; the weights were promised “by November 4” — this site's Nov 3/4 morning checks found nothing, public evidence starts with the first community mirror at 23:59 Beijing on Nov 4, and a Nov 5 re-check confirms 96 shards ≈1.56TB are public. Honored at the wire; see the site article for the reading.
Original release ↗