Public benchmarkExternal source reviewReviewed by DanPublished Source checked Updated
Kimi K3.1 Pro: public coding benchmark receipts
The report
Verdict
Published benchmark evidence
Moonshot's reported 65.7% terminal and 51.7% long-horizon results are useful task-specific references, but selected benchmark runs used Claude Code rather than DanManREAL's Kimi Code workflow.
On the tier list: The November 17 tier is Dan’s own ranking; public benchmark numbers are a separate evidence channel, not an automatic tier adjustment.
What the benchmarks report
- Terminal-Bench 4.0: 65.7%. Kimi K3.1 Pro; Claude Code harness. Reported by Moonshot AI model-card table.
- SWE-Marathon v1.1: 51.7%. Kimi K3.1 Pro; max effort; long-horizon coding. Reported by Moonshot AI model-card table.
Original sources
Source 1 — Terminal-Bench 4.0, SWE-Marathon v1.1.
What this evidence tells a reader
The published Terminal-Bench 4.0 score of 65.7% names a Claude Code harness; the SWE-Marathon v1.1 score of 51.7% concerns long-horizon software engineering at max effort. Those percentages are not competing measurements of the same task. They also say little about graphics, video timing or Blender asset quality without separate task-specific evidence.
DanManREAL has a separate firsthand Kimi K3.1 Pro field report involving a motion-graphics assignment under Kimi Code. That work records a different application, harness and deliverable. Linking the two is useful to show why benchmark performance and actual creative outcomes should be inspected independently, not to claim the benchmark predicted the result.
For someone choosing Kimi for production, the helpful question is whether the exact model and chosen terminal app can complete an artifact end to end with acceptable correction time. A successful code-benchmark percentage does not establish a reliable video-production pipeline.
How to use these results
- Check whether a vendor's benchmark used Claude Code while your workflow uses Kimi Code.
- For creative production, inspect actual rendered output and required deliverables rather than a software score.
- Use the linked firsthand field report for its motion task; assess coding reliability on a coding task separately.
Limits of this report
- The Moonshot card supplies published scores, not raw DanManREAL terminal logs or repeatability intervals.
- Results from Claude Code and Kimi Code harnesses cannot be transferred directly to one another.
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
- Terminal-Bench 4.0
- 65.7%Moonshot AI model-card table (model publisher). Kimi K3.1 Pro; Claude Code harness. Not independently rerun by DanManREAL.
- SWE-Marathon v1.1
- 51.7%Moonshot AI model-card table (model publisher). Kimi K3.1 Pro; max effort; long-horizon coding. Not independently rerun by DanManREAL.
- Source review
- November 17, 2026Date these source claims were reviewed; not a claim about when the published benchmark was executed.
Verdict
Published benchmark evidence
Tier list The November 17 tier is Dan’s own ranking; public benchmark numbers are a separate evidence channel, not an automatic tier adjustment.
danmanreal.com/tests/kimi-k3-1-pro-public-benchmarks
Models covered by external benchmarks
On the tier list's bench
- Kimi K3.1 Pro · 2 receipts on file
Related
- AI Coding Agents Tier List
- Kimi K3.1 Pro built my animation family, after a quota stop and 21 minutes of failed requests
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