Kimi K3
moonshotai· kimi verifiedMoonshot: long-context reasoning, verified on Supermagin
Kimi K3, tested through the Studio via the TokenRouter gateway. Real published run with a re-verifiable attestation: run the official eval to refresh the score.
2 attested runs · pooled from 1 workspace
Supermagin Verified badge· free for verified models
auto-updates with the verified score[](https://supermagin.site/public/models/openai/kimi-k3-free)Paste the line above into your GitHub repo README or Hugging Face model card. The badge only renders for models with real published runs on the Supermagin leaderboard.
Model card & documentation
Kimi K3
Moonshot’s long-context reasoning model, verified on Supermagin via a real published run through the TokenRouter gateway (moonshotai/kimi-k3-free).
Verification
- Evidence: verified on Supermagin: the score comes from a real published run, re-verifiable via its attestation hash.
- Gateway: OpenAI-compatible, base URL
https://api.tokenrouter.com/v1. - Model id:
moonshotai/kimi-k3-free(themoonshotai/prefix is required by the gateway).
Pooled score
0%
credibility 38% · consistency 0.93
Score trend across attested runs
6 total samples
Success rate
0%
Failure rate
0%
Hallucination
0%
Avg latency
64588ms
per task
Avg cost
$0.0022
per task
Calibration gap
+0.00
positive = overconfident
Scores by category
Pooled brain diagnostics
Failure rate by task type
Attestation-verified runs
Every number is pooled from opt-in published runs and re-verifiable via its attestation hash · updates live.