Kimi K3 (high)vsLuna 5.6 (max reasoning)
1 prompt where both models ran the exact same instructions. Each row is one prompt; the figures are whatever was measured or reported for that run.
Kimi K3 (high) and Luna 5.6 (max reasoning) ran the same prompt on 1 task, side by side. Cost, duration and outcome for each — 0 measured locally, no aggregate score.
- Shared prompts
- 1
- Compared
- models
- Measured runs
- 0
No winner is declared. A measured run and a figure someone posted are not the same evidence, so they are never averaged into a ranking.
What this page shows, and how to read it
Kimi K3 (high) is a large language model from Moonshot AI. On the prompts below, it ran through Pi. Luna 5.6 (max reasoning) is a large language model from OpenAI. On the prompts below, it ran through Pi.
The two sides share 1 prompt on this site: "Minecraft — five models build it from a two-line prompt". It is a games prompt, rated brutal, published by Matt on Aug 1, 2026, tagged voxel, 3d, webgl, browser, clone and one-prompt. Every run has a video capture on the prompt page. Across it, Kimi K3 (high) completed its run, and Luna 5.6 (max reasoning) completed its run. Kimi K3 (high) has a reported cost of $2.87. Luna 5.6 (max reasoning) has a reported cost of $0.35.
How to read the outcomes. Completed means the prompt alone produced a working artefact. Failed means this run did not — the build broke, the result would not run, or the agent stopped short of a working state. That is a fact about one run on one prompt, not a verdict on the tool: the same programs complete other prompts elsewhere on this site. Timed out means the run was cut at its time limit with the artefact unfinished, so whatever it cost bought a partial result. Stopped and Interrupted mark runs ended from the outside before they concluded. A low cost attached to a run that did not finish is not a saving — it is the price of an attempt, which is why every figure on this page travels with its status.
Every figure above carries a trust label. Measured means SamePrompt ran it locally in Bench Arena, reconciled the tokens on the harness log and recomputed the cost from them. Reported means the author of the source announced the figure; it is shown as stated and cannot be verified here. The two are never summed, averaged or ranked, and this page declares no winner: a cheap run that failed and an expensive run that completed are two facts, not a score.