Compare / DeepSeek V3.1 Terminus vs Mistral Large 3 2512

DeepSeek V3.1 Terminus vs Mistral Large 3 2512

20 results from 3 sources that measured both models. Each row is in the source's own units; there is no overall winner.

DeepSeek V3.1 Terminus
DeepSeek · profile
Mistral Large 3 2512
Mistral · profile
Results compared
20
Shared sources
3
BenchmarkMetricDeepSeek V3.1 TerminusMistral Large 3 2512
Arena (formerly LMArena)reportedBusiness, management and financeArena rating 1,4231,409
Arena (formerly LMArena)reportedCreative writingArena rating 1,4051,373
Arena (formerly LMArena)reportedInstruction followingArena rating 1,4201,400
Arena (formerly LMArena)reportedOverallArena rating 1,4171,413
Arena (formerly LMArena)reportedWriting, literature and languageArena rating 1,4111,388
Artificial AnalysisreportedArtificial Analysis Coding Indexindex score 43.520.1
Artificial AnalysisreportedArtificial Analysis Intelligence Indexindex score 14.89.30
Artificial AnalysisreportedGPQA Diamond% of questions 79.2%68.0%
Artificial AnalysisreportedHumanity's Last Exam% of questions 16.4%4.2%
Artificial AnalysisreportedIFBench% of instructions 57.0%36.2%
Artificial AnalysisreportedLong-context reasoning (AA-LCR)% of questions 69.3%36.0%
Artificial AnalysisreportedSciCode% of problems 38.0%36.6%
Artificial AnalysisreportedTerminal-Bench 2.1% of tasks 44.9%12.0%
Artificial AnalysisreportedTerminal-Bench 4.0% of tasks 0.0%0.0%
Artificial AnalysisreportedTerminal-Bench Hard% of tasks 31.8%15.9%
Artificial AnalysisreportedΤ-bench banking% of tasks 21.0%5.8%
Artificial AnalysisreportedΤ²-bench telecom% of tasks 37.1%24.6%
UGI LeaderboardreportedRequested-length error% off the requested word count 32.0%15.0%
UGI LeaderboardreportedStyle adherencescore from 0 to 1 0.370.38
UGI LeaderboardreportedWriting scorescore out of 100 54.041.6

Bold green marks the better value on that metric. Where a source reports ranges that overlap, the difference may not be meaningful; see the benchmark page for ranges.