Field testTested AIBy DanPublished Tested Updated
Kimi K3.1 Pro built my animation family, after a quota stop and 21 minutes of failed requests
The report
Verdict
Useful output, rough ride
Kimi built a reusable motion family and corrected wide and tall videos that I called pretty good. Getting there took a quota stop, a manual resume and nine failed requests that ate 40.73% of the resumed turn.
On the tier list: One task with no matched rival run, so it ranks nothing. When published November 7, 2026, Kimi was in S tier. The owner placed Kimi in A tier on November 17. Neither placement is a score produced by this report.
The job
On October 24 I had Kimi K3.1 Pro, at max thinking in the Kimi Code app, build a motion-graphics family for a narrated scene in my video factory. The scene explains a process: find an agent, check who it is, carry the choice forward and ask for help. Kimi built a family that handles 1 to 8 stages and reuses the same cues for wide and tall layouts.
What went wrong on the way
An earlier setup session spent much of its 61 requests on a contract for a DeepSeek helper and paused before building anything. In the main run Kimi did real work for about 27 minutes, then hit its 5-hour quota. After I resumed it, nine requests failed and the waits added up to 21 minutes 9 seconds. It also said a tall layout was clear while a crop still showed a collision, and fixed it in later versions.
The usage meter
I logged the 5-hour meter at 30%, 48.96% and 77.73% after the reset. An early guess at how the quota is counted did not fit those later readings, so I make no claim about Kimi's quota formula, or that cached input is free.
My verdict
When I published this report on November 7, Kimi K3.1 Pro was in S tier as my review specialist. My November 17 update placed it in A. This run still shows it can build too. Expect to babysit it through setup loops and failed requests.
Limits of this report
- One motion-graphics task, not a 3D, photorealism or full-episode test.
- No matched Gemini 4 Argon run on this task, so no winner is named.
- Token counts are request-log counters, not subscription charges or dollars.
- The cause of the failed requests is unverified; my modem was also reconnected during the outage.
- The new files were not adopted into the factory's production baseline.
From the channel
Follow the lab
New test reports and tier-list changes get announced in my Discord's #news channel. The full videos are on YouTube.
Receipts
DanManREAL · AI lab
Test receipt
Kimi K3.1 Pro
- Output
- 2 motion videos of 28.5 s, plus 6 source, test and doc fileswide and tall composed separately; kept in an isolated worktree
- My verdict on the corrected videos
- "The animations are pretty good."
- Quota stop
- 5-hour usage at 100%monthly display 5.50%; resumed after the reset
- Failed requests after the resume
- 9, totalling 21 min 9 s40.73% of the 51 min 56 s resumed turn; cause unverified
- Whole run, with its reviewer helper
- 106 requests, 268 tool callsabout 90 minutes wall clock, including the quota gap
- Tokens, whole run with helper
- 546,541 uncached in · 22,389,256 cached in · 126,090 outcounters from the request log, not a bill
- Earlier setup session
- 61 requests in 13 min 34 s27 went to a contract for a DeepSeek helper, with 5 rejected attempts
Verdict
Useful output, rough ride
Tier list One task with no matched rival run, so it ranks nothing. When published November 7, 2026, Kimi was in S tier. The owner placed Kimi in A tier on November 17. Neither placement is a score produced by this report.
danmanreal.com/tests/kimi-k3-1-pro-motion-family-field-test
What was tested
On the tier list's bench
- Kimi K3.1 Pro · 2 receipts on file
Shorts
Related
Independent testing. Sponsorships are always disclosed, and verdicts are never for sale.
Disagree? Bring your receiptsAll reports
- One inbox for my AI coding agents: triage took 34% less time, resuming work did not improveHead-to-headNov 20
- Mythos 5.2: research and security benchmark source reviewPublic benchmarkNov 17
- Fable 5.2: agentic coding benchmark source reviewPublic benchmarkNov 17
- Claude Opus 5.5: public coding benchmark receiptsPublic benchmarkNov 17
- Astra 6.2: new model and public task-evaluation evidencePublic benchmarkNov 17
- Sol 6.2: new model and public coding benchmark evidencePublic benchmarkNov 17
- Gemini 4 Carbon: benchmark source and usage profilePublic benchmarkNov 17
- Qwen 4: Qwen4-Max launch and benchmark sourcesPublic benchmarkNov 17
- Kimi K3.1 Pro: public coding benchmark receiptsPublic benchmarkNov 17
- GLM-5.3: public coding benchmark receiptsPublic benchmarkNov 17
- Gemini 4 Argon: public coding benchmark receiptsPublic benchmarkNov 17
- Claude Sonnet 5.5: public coding benchmark receiptsPublic benchmarkNov 17
- DeepSeek V4.1 Flash: public coding benchmark receiptsPublic benchmarkNov 17
- Muse Wonder Full 1.0: public coding benchmark receiptsPublic benchmarkNov 17
- Grok 4.7: separate release and public coding resultsPublic benchmarkNov 17
- ai-design-check on two of my FABLE screens: 10 warnings, 4 of them false alarmsField testNov 2
- Vercel Agent Desktop used a sample Trip app in 6 actions without moving the mouseField testNov 1
- Cerberus Guard on sample data: one crawler name, two verdicts, and a block earned by behaviorField testOct 31
- GPT-6.1 Terra vs Opus 5.5: three real projects, and the designs I picked blindHead-to-headOct 29
- Steps → Reduced: opening a terminal agent went from 98 seconds to 11AuditOct 28
- One ChatGPT chat directed Opus and Codex at once through DMR_REMOTEField testOct 28
- Claude Duet ran my channel for a week and caught 269 fake AI crawlersField testOct 27
- Kimi K3.1 Pro built my animation family, after a quota stop and 21 minutes of failed requestsField testOct 24
- ChatGPT + Codex built 33 new DMR_REMOTE toolsField testOct 23
- Gemini 4 Argon reviewed a finished episode: 14 clips, one flag, one inconsistencyField testOct 21
- Qwen 3.8 Flash Next on my RTX 5080: 81 tokens a second with 202K tokens readAuditOct 21
- Does an extra AI review step pay off? A small DeepSeek trial said noHead-to-headOct 19
- My real DeepSeek bill: 12.31 billion tokens for $122.28Cost receiptOct 17
- Gemini 4 Argon kept hanging in Gemini CLI. The cause was a 60-second timeoutIncidentOct 14
- I asked 10 AI coding agents for their biggest weaknessSelf-reportOct 14
- Jarvis 27B on my own PC: 30 tokens a second, and it still recalls at 251KAuditOct 10
- My agents' first game benchmark was wrong. Here's how it was caughtIncidentOct 10
- Pixal3D + Blender turned my 2D art into animated 3D charactersField testSep 22
- DeepSeek Flash answered 200 OK and sent nothing for two hoursIncidentSep 14
- Grok, Codex and DeepSeek played WorldOS for Rank 1. Grok won with 98Head-to-headAug 20
- DeepSeek V4 Flash audited 18 of my real coding sessions. Two said 'done' too earlyAuditAug 4