Kimi K3 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
The model did not finish, so there is no output.
Wall time 16m 35sCost $0.52 reported
Review
No verdict yet.
The full review
DNF. Harness exited after 16.6 min without writing index.html. The model's last message stopped at the output-token limit (stop reason max_tokens / length) without a tool call, so the non-interactive run ended there.
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, 997 characters. Every result for this prompt
Make a complete playable game in ONE html file under 13 KB (minified size reported in a corner): the visuals are rendered entirely by one fragment shader (raymarched or 2D SDF). Arcade loop: dodge/collect, score, increasing speed, game over + restart. Keyboard and touch controls. No libraries, no assets. IDLE AUTO-PLAY: with no input, the page starts animating immediately and runs an automatic demo (camera motion, preset or seed cycling, or an AI player) so a headless capture shows constant motion. Must render on WebGL/WebGL2 with no WebGPU required. Save the main page as index.html in the current directory (plus any local asset files you create), with no build step. You may serve it locally and open it in headless chromium (/usr/bin/chromium, software WebGL, no GPU, no WebGPU) 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. Finish with a short README.md describing the controls.
How it was made
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 2 steps
Loading the timeline…
Download the full transcript OpenCode session export, 99 KB. 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 $0.52, as reported by OpenCode (its own bill for the run). The list-price estimate from the token counts is $0.52.
| Tokens by type | Tokens | $ per 1M | Cost |
|---|---|---|---|
| Input (uncached) | 13,938 | $3.00 | $0.0418 |
| Cache write | 0 | $3.00 | $0 |
| Cache read | 0 | $0.30 | $0 |
| Output | 32,000 | $15.00 | $0.4800 |
| of which reasoning | 32,000 | billed as output |
- Total computed from tokens
- $0.5218
- Total reported by the harness
- $0.5218
- Difference
- matches
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 (0 visible + 32,000 reasoning), and both bill at the output rate.
Prices: https://platform.kimi.ai/docs/pricing/chat, 2026-09-24.
Run notes
- frozen prompt (written): core requirements sha256 0217f6e8b2ea, full harness prompt sha256 0217f6e8b2ea; exact prompt text used by the existing harness rows of this test on the runs branch (copied from js13k_style_shader_game__claude-code__haiku-5-5__attempt1/meta.json, sha256 verified); Beta half of the
- 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
- Skills on: the run had 54 skills installed (33 video + 21 web/Three.js), as its meta records.
- Kimi K3 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.