Decision 2.0: local benchmark

Updated 2026-10-05T00:49:41.843395+00:00

231 public JevBench cases; original scoring. Separate from the sealed leaderboard and from full English/Norwegian evaluation. Results use the same pinned inputs and retain native rejections as failures. Scores appear only when the complete suite has been attempted.

All four native models completed all five frozen fixtures: 70,304 saved case predictions, 0 final saved prediction errors. Windows allocation/kernel failures interrupted 4 process attempts; exact-runtime retries resumed saved IDs. Independent result verification and crash history.

Public JevBench comparison

ModelCases attemptedOriginal-score accuracyInference errors
Decision 2.0 Kai 0.6B231/23165.8%0
Decision 2.0 Eos 0.8B231/23170.6%0
Decision 2.0 Sol 2B231/23171.9%0
Decision 2.0 Nox 4B231/23180.1%0
Decision 4B v1.2231/23187.9%0

Decision 4B v1.2 is the separate FlyMy model previously benchmarked here.

Completed shared suites

ModelSuiteCasesExact-match accuracyCoverage
Decision 2.0 Kai 0.6Bjev_verified/agnews10083.0%100.0%
Decision 2.0 Kai 0.6Bjev_verified/emotiondair10053.0%100.0%
Decision 2.0 Kai 0.6Bjev_verified/banking7710081.0%100.0%
Decision 2.0 Kai 0.6Bfresh/news20083.5%100.0%
Decision 2.0 Kai 0.6Bfresh/emotion20057.0%100.0%
Decision 2.0 Kai 0.6Bfresh/spam20054.5%100.0%
Decision 2.0 Kai 0.6Bfresh/phishing20068.5%100.0%
Decision 2.0 Kai 0.6Bfresh_business/routing/en1258.3%100.0%
Decision 2.0 Kai 0.6Bfresh_business/routing/nb1216.7%100.0%
Decision 2.0 Kai 0.6Bfresh_business/documents/en1291.7%100.0%
Decision 2.0 Kai 0.6Bfresh_business/documents/nb1291.7%100.0%
Decision 2.0 Kai 0.6Bindustry/routing/en4087.5%100.0%
Decision 2.0 Kai 0.6Bindustry/routing/en/reversed4087.5%100.0%
Decision 2.0 Kai 0.6Bindustry/routing/nb4077.5%100.0%
Decision 2.0 Kai 0.6Bindustry/routing/nb/reversed4077.5%100.0%
Decision 2.0 Kai 0.6Bindustry/documents/en28100.0%100.0%
Decision 2.0 Kai 0.6Bindustry/documents/en/reversed28100.0%100.0%
Decision 2.0 Kai 0.6Bindustry/documents/nb2892.9%100.0%
Decision 2.0 Kai 0.6Bindustry/documents/nb/reversed2896.4%100.0%
Decision 2.0 Kai 0.6Bjev_native/agnews10082.0%100.0%
Decision 2.0 Kai 0.6Bjev_native/emotiondair10053.0%100.0%
Decision 2.0 Kai 0.6Bjev_native/banking7710081.0%100.0%
Decision 2.0 Kai 0.6Bfast_decisions_dev/agent_handoff10058.0%100.0%
Decision 2.0 Kai 0.6Bfast_decisions_dev/banking_intent10055.0%100.0%
Decision 2.0 Kai 0.6Bfast_decisions_dev/benefits_request10049.0%100.0%
Decision 2.0 Kai 0.6Bfast_decisions_dev/clinic_request10041.0%100.0%
Decision 2.0 Kai 0.6Bfast_decisions_dev/document_type10080.0%100.0%
Decision 2.0 Kai 0.6Bfast_decisions_dev/email_triage1008.0%100.0%
Decision 2.0 Kai 0.6Bfast_decisions_dev/news_topic10057.0%100.0%
Decision 2.0 Kai 0.6Bfast_decisions_dev/paper_field10042.0%100.0%
Decision 2.0 Kai 0.6Bfast_decisions_dev/product_feedback1004.0%100.0%
Decision 2.0 Kai 0.6Bfast_decisions_dev/restaurant_review10013.0%100.0%
Decision 2.0 Kai 0.6Bfast_decisions_dev/review_sentiment10068.0%100.0%
Decision 2.0 Kai 0.6Bfast_decisions_dev/screen_tags10015.0%100.0%
Decision 2.0 Kai 0.6Bfast_decisions_dev/sports_recap10040.0%100.0%
Decision 2.0 Kai 0.6Bfast_decisions_dev/support_intent10060.0%100.0%
Decision 2.0 Kai 0.6Bfast_decisions_dev/support_topic10040.0%100.0%
Decision 2.0 Kai 0.6Bfast_decisions_dev/ticket_route1004.0%100.0%
Decision 2.0 Kai 0.6Bfast_decisions_dev/travel_request10060.0%100.0%
Decision 2.0 Kai 0.6Blaya_claim/news.replication40083.8%100.0%
Decision 2.0 Kai 0.6Blaya_claim/emotion.replication40055.0%100.0%
Decision 2.0 Kai 0.6Blaya_claim/spam.replication40048.0%100.0%
Decision 2.0 Kai 0.6Blaya_claim/phishing.replication40071.5%100.0%
Decision 2.0 Kai 0.6Blaya_claim/jailbreaking40087.5%100.0%
Decision 2.0 Kai 0.6Blaya_claim/toxicity40065.8%100.0%
Decision 2.0 Kai 0.6Blaya_claim/intent30085.7%100.0%
Decision 2.0 Kai 0.6Blaya_claim/scenario30066.3%100.0%
Decision 2.0 Kai 0.6Bmassive_en294872.9%100.0%
Decision 2.0 Kai 0.6Bmassive_nb294854.7%100.0%
Decision 2.0 Kai 0.6Bnorec_no117357.7%100.0%
Decision 2.0 Kai 0.6Bintern_claim/typed_decisions-test4003.8%100.0%
Decision 2.0 Eos 0.8Bjev_verified/agnews10073.0%100.0%
Decision 2.0 Eos 0.8Bjev_verified/emotiondair10049.0%100.0%
Decision 2.0 Eos 0.8Bjev_verified/banking7710090.0%100.0%
Decision 2.0 Eos 0.8Bfresh/news20082.5%100.0%
Decision 2.0 Eos 0.8Bfresh/emotion20051.0%100.0%
Decision 2.0 Eos 0.8Bfresh/spam20069.5%100.0%
Decision 2.0 Eos 0.8Bfresh/phishing20076.0%100.0%
Decision 2.0 Eos 0.8Bfresh_business/routing/en1250.0%100.0%
Decision 2.0 Eos 0.8Bfresh_business/routing/nb1241.7%100.0%
Decision 2.0 Eos 0.8Bfresh_business/documents/en1283.3%100.0%
Decision 2.0 Eos 0.8Bfresh_business/documents/nb1283.3%100.0%
Decision 2.0 Eos 0.8Bindustry/routing/en4085.0%100.0%
Decision 2.0 Eos 0.8Bindustry/routing/en/reversed4082.5%100.0%
Decision 2.0 Eos 0.8Bindustry/routing/nb4065.0%100.0%
Decision 2.0 Eos 0.8Bindustry/routing/nb/reversed4067.5%100.0%
Decision 2.0 Eos 0.8Bindustry/documents/en28100.0%100.0%
Decision 2.0 Eos 0.8Bindustry/documents/en/reversed28100.0%100.0%
Decision 2.0 Eos 0.8Bindustry/documents/nb2889.3%100.0%
Decision 2.0 Eos 0.8Bindustry/documents/nb/reversed2889.3%100.0%
Decision 2.0 Eos 0.8Bjev_native/agnews10082.0%100.0%
Decision 2.0 Eos 0.8Bjev_native/emotiondair10049.0%100.0%
Decision 2.0 Eos 0.8Bjev_native/banking7710084.0%100.0%
Decision 2.0 Eos 0.8Bfast_decisions_dev/agent_handoff10065.0%100.0%
Decision 2.0 Eos 0.8Bfast_decisions_dev/banking_intent10054.0%100.0%
Decision 2.0 Eos 0.8Bfast_decisions_dev/benefits_request10044.0%100.0%
Decision 2.0 Eos 0.8Bfast_decisions_dev/clinic_request10041.0%100.0%
Decision 2.0 Eos 0.8Bfast_decisions_dev/document_type10076.0%100.0%
Decision 2.0 Eos 0.8Bfast_decisions_dev/email_triage1005.0%100.0%
Decision 2.0 Eos 0.8Bfast_decisions_dev/news_topic10056.0%100.0%
Decision 2.0 Eos 0.8Bfast_decisions_dev/paper_field10035.0%100.0%
Decision 2.0 Eos 0.8Bfast_decisions_dev/product_feedback1004.0%100.0%
Decision 2.0 Eos 0.8Bfast_decisions_dev/restaurant_review10011.0%100.0%
Decision 2.0 Eos 0.8Bfast_decisions_dev/review_sentiment10069.0%100.0%
Decision 2.0 Eos 0.8Bfast_decisions_dev/screen_tags10019.0%100.0%
Decision 2.0 Eos 0.8Bfast_decisions_dev/sports_recap10041.0%100.0%
Decision 2.0 Eos 0.8Bfast_decisions_dev/support_intent10058.0%100.0%
Decision 2.0 Eos 0.8Bfast_decisions_dev/support_topic10039.0%100.0%
Decision 2.0 Eos 0.8Bfast_decisions_dev/ticket_route1005.0%100.0%
Decision 2.0 Eos 0.8Bfast_decisions_dev/travel_request10067.0%100.0%
Decision 2.0 Eos 0.8Blaya_claim/news.replication40079.0%100.0%
Decision 2.0 Eos 0.8Blaya_claim/emotion.replication40052.8%100.0%
Decision 2.0 Eos 0.8Blaya_claim/spam.replication40075.0%100.0%
Decision 2.0 Eos 0.8Blaya_claim/phishing.replication40076.5%100.0%
Decision 2.0 Eos 0.8Blaya_claim/jailbreaking40092.5%100.0%
Decision 2.0 Eos 0.8Blaya_claim/toxicity40070.0%100.0%
Decision 2.0 Eos 0.8Blaya_claim/intent30084.7%100.0%
Decision 2.0 Eos 0.8Blaya_claim/scenario30076.7%100.0%
Decision 2.0 Eos 0.8Bmassive_en294867.5%100.0%
Decision 2.0 Eos 0.8Bmassive_nb294852.6%100.0%
Decision 2.0 Eos 0.8Bnorec_no117353.4%100.0%
Decision 2.0 Eos 0.8Bintern_claim/typed_decisions-test4002.0%100.0%
Decision 2.0 Sol 2Bjev_verified/agnews10080.0%100.0%
Decision 2.0 Sol 2Bjev_verified/emotiondair10052.0%100.0%
Decision 2.0 Sol 2Bjev_verified/banking7710089.0%100.0%
Decision 2.0 Sol 2Bfresh/news20083.5%100.0%
Decision 2.0 Sol 2Bfresh/emotion20029.0%100.0%
Decision 2.0 Sol 2Bfresh/spam20074.5%100.0%
Decision 2.0 Sol 2Bfresh/phishing20086.0%100.0%
Decision 2.0 Sol 2Bfresh_business/routing/en1258.3%100.0%
Decision 2.0 Sol 2Bfresh_business/routing/nb1258.3%100.0%
Decision 2.0 Sol 2Bfresh_business/documents/en12100.0%100.0%
Decision 2.0 Sol 2Bfresh_business/documents/nb12100.0%100.0%
Decision 2.0 Sol 2Bindustry/routing/en4087.5%100.0%
Decision 2.0 Sol 2Bindustry/routing/en/reversed4077.5%100.0%
Decision 2.0 Sol 2Bindustry/routing/nb4075.0%100.0%
Decision 2.0 Sol 2Bindustry/routing/nb/reversed4070.0%100.0%
Decision 2.0 Sol 2Bindustry/documents/en28100.0%100.0%
Decision 2.0 Sol 2Bindustry/documents/en/reversed2896.4%100.0%
Decision 2.0 Sol 2Bindustry/documents/nb28100.0%100.0%
Decision 2.0 Sol 2Bindustry/documents/nb/reversed2896.4%100.0%
Decision 2.0 Sol 2Bjev_native/agnews10080.0%100.0%
Decision 2.0 Sol 2Bjev_native/emotiondair10049.0%100.0%
Decision 2.0 Sol 2Bjev_native/banking7710090.0%100.0%
Decision 2.0 Sol 2Bfast_decisions_dev/agent_handoff10075.0%100.0%
Decision 2.0 Sol 2Bfast_decisions_dev/banking_intent10055.0%100.0%
Decision 2.0 Sol 2Bfast_decisions_dev/benefits_request10050.0%100.0%
Decision 2.0 Sol 2Bfast_decisions_dev/clinic_request10038.0%100.0%
Decision 2.0 Sol 2Bfast_decisions_dev/document_type10080.0%100.0%
Decision 2.0 Sol 2Bfast_decisions_dev/email_triage1007.0%100.0%
Decision 2.0 Sol 2Bfast_decisions_dev/news_topic10058.0%100.0%
Decision 2.0 Sol 2Bfast_decisions_dev/paper_field10041.0%100.0%
Decision 2.0 Sol 2Bfast_decisions_dev/product_feedback1003.0%100.0%
Decision 2.0 Sol 2Bfast_decisions_dev/restaurant_review10019.0%100.0%
Decision 2.0 Sol 2Bfast_decisions_dev/review_sentiment10076.0%100.0%
Decision 2.0 Sol 2Bfast_decisions_dev/screen_tags10018.0%100.0%
Decision 2.0 Sol 2Bfast_decisions_dev/sports_recap10044.0%100.0%
Decision 2.0 Sol 2Bfast_decisions_dev/support_intent10055.0%100.0%
Decision 2.0 Sol 2Bfast_decisions_dev/support_topic10049.0%100.0%
Decision 2.0 Sol 2Bfast_decisions_dev/ticket_route10011.0%100.0%
Decision 2.0 Sol 2Bfast_decisions_dev/travel_request10066.0%100.0%
Decision 2.0 Sol 2Blaya_claim/news.replication40082.0%100.0%
Decision 2.0 Sol 2Blaya_claim/emotion.replication40031.5%100.0%
Decision 2.0 Sol 2Blaya_claim/spam.replication40080.0%100.0%
Decision 2.0 Sol 2Blaya_claim/phishing.replication40085.2%100.0%
Decision 2.0 Sol 2Blaya_claim/jailbreaking40091.8%100.0%
Decision 2.0 Sol 2Blaya_claim/toxicity40079.0%100.0%
Decision 2.0 Sol 2Blaya_claim/intent30084.7%100.0%
Decision 2.0 Sol 2Blaya_claim/scenario30072.3%100.0%
Decision 2.0 Sol 2Bmassive_en294873.4%100.0%
Decision 2.0 Sol 2Bmassive_nb294864.4%100.0%
Decision 2.0 Sol 2Bnorec_no117357.2%100.0%
Decision 2.0 Sol 2Bintern_claim/typed_decisions-test4006.2%100.0%
Decision 2.0 Nox 4Bjev_verified/agnews10085.0%100.0%
Decision 2.0 Nox 4Bjev_verified/emotiondair10051.0%100.0%
Decision 2.0 Nox 4Bjev_verified/banking7710086.0%100.0%
Decision 2.0 Nox 4Bfresh/news20087.0%100.0%
Decision 2.0 Nox 4Bfresh/emotion20059.0%100.0%
Decision 2.0 Nox 4Bfresh/spam20093.0%100.0%
Decision 2.0 Nox 4Bfresh/phishing20092.5%100.0%
Decision 2.0 Nox 4Bfresh_business/routing/en1283.3%100.0%
Decision 2.0 Nox 4Bfresh_business/routing/nb1283.3%100.0%
Decision 2.0 Nox 4Bfresh_business/documents/en1291.7%100.0%
Decision 2.0 Nox 4Bfresh_business/documents/nb12100.0%100.0%
Decision 2.0 Nox 4Bindustry/routing/en4095.0%100.0%
Decision 2.0 Nox 4Bindustry/routing/en/reversed4095.0%100.0%
Decision 2.0 Nox 4Bindustry/routing/nb4087.5%100.0%
Decision 2.0 Nox 4Bindustry/routing/nb/reversed4085.0%100.0%
Decision 2.0 Nox 4Bindustry/documents/en28100.0%100.0%
Decision 2.0 Nox 4Bindustry/documents/en/reversed2896.4%100.0%
Decision 2.0 Nox 4Bindustry/documents/nb28100.0%100.0%
Decision 2.0 Nox 4Bindustry/documents/nb/reversed28100.0%100.0%
Decision 2.0 Nox 4Bjev_native/agnews10088.0%100.0%
Decision 2.0 Nox 4Bjev_native/emotiondair10053.0%100.0%
Decision 2.0 Nox 4Bjev_native/banking7710087.0%100.0%
Decision 2.0 Nox 4Bfast_decisions_dev/agent_handoff10081.0%100.0%
Decision 2.0 Nox 4Bfast_decisions_dev/banking_intent10059.0%100.0%
Decision 2.0 Nox 4Bfast_decisions_dev/benefits_request10047.0%100.0%
Decision 2.0 Nox 4Bfast_decisions_dev/clinic_request10048.0%100.0%
Decision 2.0 Nox 4Bfast_decisions_dev/document_type10080.0%100.0%
Decision 2.0 Nox 4Bfast_decisions_dev/email_triage1008.0%100.0%
Decision 2.0 Nox 4Bfast_decisions_dev/news_topic10067.0%100.0%
Decision 2.0 Nox 4Bfast_decisions_dev/paper_field10043.0%100.0%
Decision 2.0 Nox 4Bfast_decisions_dev/product_feedback10013.0%100.0%
Decision 2.0 Nox 4Bfast_decisions_dev/restaurant_review10042.0%100.0%
Decision 2.0 Nox 4Bfast_decisions_dev/review_sentiment10087.0%100.0%
Decision 2.0 Nox 4Bfast_decisions_dev/screen_tags10019.0%100.0%
Decision 2.0 Nox 4Bfast_decisions_dev/sports_recap10048.0%100.0%
Decision 2.0 Nox 4Bfast_decisions_dev/support_intent10069.0%100.0%
Decision 2.0 Nox 4Bfast_decisions_dev/support_topic10044.0%100.0%
Decision 2.0 Nox 4Bfast_decisions_dev/ticket_route10012.0%100.0%
Decision 2.0 Nox 4Bfast_decisions_dev/travel_request10067.0%100.0%
Decision 2.0 Nox 4Blaya_claim/news.replication40085.5%100.0%
Decision 2.0 Nox 4Blaya_claim/emotion.replication40058.0%100.0%
Decision 2.0 Nox 4Blaya_claim/spam.replication40093.5%100.0%
Decision 2.0 Nox 4Blaya_claim/phishing.replication40092.0%100.0%
Decision 2.0 Nox 4Blaya_claim/jailbreaking40078.8%100.0%
Decision 2.0 Nox 4Blaya_claim/toxicity40084.0%100.0%
Decision 2.0 Nox 4Blaya_claim/intent30090.3%100.0%
Decision 2.0 Nox 4Blaya_claim/scenario30075.0%100.0%
Decision 2.0 Nox 4Bmassive_en294885.2%100.0%
Decision 2.0 Nox 4Bmassive_nb294880.3%100.0%
Decision 2.0 Nox 4Bnorec_no117372.0%100.0%
Decision 2.0 Nox 4Bintern_claim/typed_decisions-test40019.0%100.0%

Fresh suite accuracies exclude development records. Score classes use the modal level; native expected scores are retained in raw results. Multilabel tasks use separate native Yes/No questions. Authored Norwegian/business suites are diagnostics.

Windows NVIDIA CUDA/Transformers/FLA environment differs from publisher ROCm latency environment. Public benchmark training contamination and model-selection exposure are not independently established. No model weights, prompts or test labels were tuned during this run. Decision 2.0 Lux 9B and Vega 27B: native precision exceeds local GPU budget

Initial execution · Resumable retries · Result summary · Concurrency and vLLM compatibility · Full comparison