AI Model Reviewer

LongCat 2.5 Preview in OpenCode 1.18.32

skills off

← Open in the gallerySame prompt, all models

Poster frame of LongCat 2.5 Preview's page
Open the original page

The model's own page, run in a sandboxed frame only when you ask (a page that needs its own origin opens in a new tab on a separate demo site). Phones get a full-screen player capped at 1.5x pixel density (2x on larger screens) so it stays smooth; the original page runs uncapped. It starts in its auto-tour (T switches to play mode on a keyboard).

ErrorTidewatermax effortNeutral harness

Wall time 19m 05sCost $0 reported

Snapshot 12 of 12 saved during LongCat 2.5 Preview's run
  1. Snapshot 1 of 12 saved during LongCat 2.5 Preview's run
  2. Snapshot 3 of 12 saved during LongCat 2.5 Preview's run
  3. Snapshot 4 of 12 saved during LongCat 2.5 Preview's run
  4. Snapshot 6 of 12 saved during LongCat 2.5 Preview's run
  5. Snapshot 7 of 12 saved during LongCat 2.5 Preview's run
  6. Snapshot 9 of 12 saved during LongCat 2.5 Preview's run
  7. Snapshot 10 of 12 saved during LongCat 2.5 Preview's run
  8. Snapshot 12 of 12 saved during LongCat 2.5 Preview's run
Still snapshots saved during the run (no motion recording): 12 distinct frames.

Generation timelapse

Poster for generation timelapse

Generation timelapse (the agent's screen during the run, sped up). This is not the finished video. Download video (403 KB).

Review

No verdict yet.

The full review

Error. The harness ended the turn with an error (APIError: Upstream request failed: Endpoint is unavailable.); cause unknown, not counted against the model.

Attempt 1 of 2

Every try of this model in this harness on this prompt, oldest first. A later attempt is the cell's result: attempt 2.

  1. Attempt 1 Error this run19m 05s$02026-09-26
  2. Attempt 2 Stopped by us the cell's result12m 24s$02026-09-26

Scores

Not scored: only runs that finished on their own are scored.

Downloads

Files actually published for this run. Original output, site-made previews and post-run conversions are labelled separately.

The exact prompt

as run, 1,336 characters. Every result for this prompt

Build an explorable tropical island in Three.js. The island has white sand beaches, palm trees, a green hilly interior and a small fishing village of wooden huts with thatched roofs, joined to the sea by a long wooden pier with boats moored alongside. The ocean around it is animated, with waves, foam along the shore and clear shallow water over a reef where fish swim among corals and sea plants, fading to deep blue further out. Let the player switch between modes: walk on the island, swim at the surface, dive underwater (with underwater fog and caustic light), sail a small boat around the island, and fly a free camera. Add a time-of-day panel with a slider that moves the sun from morning through sunset to night, changing the sky, lighting and water colours, with stars and lit windows in the village at night. Show the controls on screen. Everything is built in code from Three.js geometry, shaders and materials; no external models or textures. Build it as index.html (plus any local asset files you create) in the current directory. Three.js may load from a pinned CDN. You may serve it locally and open it in headless chromium (/usr/bin/chromium, software WebGL, no GPU) to screenshot and inspect your own output, and iterate on your own before finishing. Close any browser you open as soon as you've taken your screenshot.

How it was made

Wall time 19m 05sFirst artifact at 7m 26s13 saved versions

Versions saved during the run

13 intermediate outputs, in the order the model wrote them.

Step by step

The prompt, every agent turn and tool call (tool name and a one-line summary; inputs and outputs are not shown), then the final message. The full transcript is 2.6 MB; it is published as a scrubbed, compressed download rather than shown on the page.

Show all 77 steps

Loading the timeline…

Download the full transcript OpenCode session export, 61 KB, gzip. Scrubbed: anything that looks like a key, token, e-mail address or phone number is replaced, and account details and the test machine's file paths are removed.

Cost breakdown

Headline cost $0, as reported by OpenCode (its own bill for the run).

Tokens by typeTokens$ per 1MCost
Input (uncached)65,559n/an/a
Cache write0n/an/a
Cache read1,384,576n/an/a
Output41,564n/an/a
Total computed from tokens
unknown
Total reported by the harness
$0

OpenCode 1.18.32 reported its own total; token counts x list prices shown beside it. No list price for longcat-2.5-preview-free in config/prices.json; no estimate computed. The harness reports reasoning tokens separately from output; here output includes them (19,488 visible + 22,076 reasoning), and both bill at the output rate.

Run notes

one prompt, agentic harness (self-review allowed)prompt rewritten after the 2026-09-24 box losslabelled re-run (of threejs_tidewater__opencode__longcat-2.5-preview)

  • frozen prompt (rewritten): engine/prompts_rewritten/threejs_tidewater.txt: written by us from the recorded description (zero results)
  • plain OpenCode config copy (providers only; no ECC instructions/agents)
  • opencode export written to a file (not piped)
  • Mode: interactive TUI in tmux (160x48), inside bubblewrap sandbox
  • Isolation audit: outside paths seen