AI Model Reviewer

GPT-6 Astra in OpenCode 1.18.32

skills off

← Open in the gallerySame prompt, all models

Zoom, inline player and more ways to view
Poster frame of GPT-6 Astra'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. Tap inside for sound.

OKCity maps navigatormax effortNeutral harness

Wall time 22m 58sCost $5.54 reported

Review

No verdict yet.

The full review

OK. Finished on its own after 23.0 min.

Scores

AI judge

Not scored

No rubric for this prompt

Visitor votes and reviewer grades are not available yet.

AI judge
AI judge: the test runner (model claude-opus-5-5), scoring by visual review of the 1600 px render against the pre-registered rubric. No AI-judge score exists for this result: there is no rubric for this prompt, so no score is shown and none is made up.
Visitors
Public voting with a blind reveal is coming: you score the output first, then see the other scores.
Reviewer
The reviewer's own grade appears here once it is added.

Scores compare results for the same prompt only. AI-judge boards rank runs within one prompt, equal totals sharing a rank; scores from different prompts are never averaged.

Downloads

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

The exact prompt

as run, 2,069 characters. Every result for this prompt

Build a working maps app in one page in the spirit of a navigation app. Generate a fictional city procedurally from a fixed seed: a coastline and river, parks, a road network with highways, main roads and side streets (each with a speed), named streets, and about 60 named places (cafes, museums, stations, parks, hospitals) with category icons drawn in code. Render it as a crisp vector map on canvas with smooth pan and zoom (mouse wheel, drag, pinch), level-of-detail labels that appear as you zoom, and a compass and scale bar. It needs: a search box with fuzzy matching over place names and categories; tapping a place opens a card; directions between any two places computed with A* on the road graph with a choice of fastest versus shortest route, the route drawn as a bold line with estimated time and distance and a step list of turn-by-turn instructions (turn left onto River Road, and so on) derived from the geometry; a simulated drive mode where a car marker follows the route and the map rotates and follows it; and a traffic toggle that slows some roads and recomputes the route. So the page is alive without input, on load it runs a scripted demo: searches for a museum, opens its card, requests directions from the central station, shows the step list, starts drive mode, turns traffic on mid-drive so the route reroutes, then picks a new pair of places and repeats, until the user touches anything. Make it look like a premium, modern map style.

Save it as index.html in the current directory: one self-contained file with inline CSS and JavaScript only, no libraries or frameworks, and no external requests of any kind (no CDNs, web fonts, images or audio files). It must work when opened directly from disk, fit any window from a 390 px wide phone to a desktop, and run without console errors. 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 22m 58sFirst artifact at 7m 56s4 saved versions

Versions saved during the run

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

Version 4 of 4
  1. Version 1 at 8m 02s thumbnailv1 8:02
  2. Version 2 at 10m 44s thumbnailv2 10:44
  3. Version 3 at 14m 21s thumbnailv3 14:21
  4. Version 4 at 19m 50s thumbnailv4 19:50

No transcript was delivered with this result.

Cost breakdown

Headline cost $5.54, as reported by OpenCode (its own bill for the run). The list-price estimate from the token counts is $5.36.

Tokens by typeTokens$ per 1MCost
Input (uncached)114$10.00$0.0011
Cache write72,193$10.00$0.7219
Cache read2,025,148$1.00$2.0251
Output52,249$50.00$2.6124
of which reasoning12,952billed as output
Total computed from tokens
$5.3607
Total reported by the harness
$5.5412
Difference
Harness-reported $5.5412 vs list-price estimate $5.3607 (-3.3%); the harness figure is the headline.

OpenCode 1.18.32 reported its own total; token counts x list prices shown beside it. Cache writes priced at the input rate (the price table has no separate cache-write price). The harness reports reasoning tokens separately from output; here output includes them (39,297 visible + 12,952 reasoning), and both bill at the output rate.

Prices: Project price table (vendor list prices), 2026-09-22.

Run notes

one prompt, agentic harness (self-review allowed)

  • frozen prompt (written): written 2026-10-05 for the MR category expansion (owner request: 500 finished runs in every category); category app_clones; 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: none
  • Renderer: Canvas 2D (getContext('2d') at index.html line 105) plus DOM/CSS UI and inline SVG; single self-contained index.html, no CDN and no external assets.
  • Card video: the run's own session capture (headless Chromium, no GPU), re-encoded as H.264. Poster: the run's own render.png.
  • Same frozen prompt as the other rows on this test; configurations differ and all rows are unscored, so this is not a head-to-head result.