MiniMax M3 in OpenCode 1.18.35
skills on · OpenCode 1.18.35 + skills on (54: 33 video + 21 web/Three.js)Kit: 33 video + 21 web/Three.js · 0 of 54 installed skills used in the session
Zoom, inline player and more ways to view
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.
Wall time 32m 15sCost $1.58 reported
Review
No verdict yet.
The full review
OK. Harness exited after 32.3 min with the artifact written (final log event not seen).
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, 1,746 characters. Every result for this prompt
Build a pathfinding race visualizer: four copies of the same grid maze, side by side (stacked on a phone), where A*, Dijkstra, breadth-first search and greedy best-first search solve the same start-to-goal problem at the same time, step by step. Show each algorithm's frontier, visited cells and final path in distinct colours, and live counters for cells visited, path length and steps taken; when all four finish, show a small results table ranking them and say which found the shortest path and which explored the fewest cells. Controls: generate a new random maze (a perfect maze with some extra openings so paths differ), clear the grid, draw or erase walls by dragging on any grid (the edit applies to all four), move the start and goal by dragging them, a speed slider, play, pause and single step. Diagonal moves are off; every move costs 1. So the page is alive without input, on load it generates a maze and starts the race automatically, and when a race finishes it waits 3 seconds and runs a new maze, forever, until the user touches a control. Make it clear and beautiful, the kind of visual a teacher would put on a projector. 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
Versions saved during the run
12 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 machine-format transcript is published as a scrubbed download.
Show all 400 steps
Loading the timeline…
Download the full transcript OpenCode session export, 1.3 MB. 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 $1.58, as reported by OpenCode (its own bill for the run). The list-price estimate from the token counts is $1.55.
| Tokens by type | Tokens | $ per 1M | Cost |
|---|---|---|---|
| Input (uncached) | 272,182 | $0.30 | $0.0817 |
| Cache write | 0 | $0.30 | $0 |
| Cache read | 22,110,616 | $0.06 | $1.3045 |
| Output | 139,496 | $1.20 | $0.1674 |
| of which reasoning | 67,885 | billed as output |
- Total computed from tokens
- $1.5536
- Total reported by the harness
- $1.5757
- Difference
- computed $1.55 vs reported $1.58 (diff -0.02)
OpenCode 1.18.35 reported its own total; token counts x list prices shown beside it. The harness reports reasoning tokens separately from output; here output includes them (71,611 visible + 67,885 reasoning), and both bill at the output rate.
Prices: https://fireworks.ai/models/fireworks/minimax-m3, 2026-10-08.
Run notes
- frozen prompt (written): core requirements sha256 14da70349e0a, full harness prompt sha256 14da70349e0a; exact prompt text used by the existing harness rows of this test on the runs branch (copied from algoviz_pathfinding_race__claude-code__fable-5-1__attempt1/meta.json, sha256 verified); Beta half of th
- no TUI: the harness ran in its non-interactive mode, so there are no TUI screenshots; the timelapse is built from version renders only
- OpenCode config (providers only, permission allow-all except question; no agents/instructions) + both default skill kits in xdg/config/opencode/skill (skills=on)
- infra retries: 0
- in the published transcript the sandbox HOME directory is shown as <HOME>
- skills=on: 54 skills installed in the harness skill directory before launch (video-33 remotion-dev/skills + heygen-com/hyperframes@0b5db9da + f; web-threejs-21 [redacted]/ai-AI Model Reviewer skills/ (alpha/mr-homebar)@7521074d98c9).
- Mode: headless run mode (opencode run --format json), inside bubblewrap sandbox
- Isolation audit: none
- Renderer: canvas 2D (the maze race is drawn on a <canvas> with getContext('2d') by the page's own script); no library, nothing loaded from a CDN (single self-contained file).
- Card video: the run's own video.mp4 (H.264 yuv420p, unchanged). Poster: the run's own render.png.
- Skills on: the run had 54 skills installed (33 video + 21 web/Three.js), as its meta records.
- MiniMax M3 served via Fireworks (OpenCode provider fireworks-ai).
- 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.