AI Model Timeline

Tracking the accelerating release frequency of frontier AI models.

DeepSeek-Terminus

DeepSeek·

LLMsopen-weightcloud + local

Overview

Experimental model pushing the boundaries of open weights.

Capabilities and innovations

Experimental archCommunity researchNovel attention

Benchmarks

BenchmarkScoreSource
GPQA Diamond83.7%unsourced
Humanity’s Last Exam (no tools)29.5%unsourced
SWE-bench Verified58.3%unsourced
MMLU89.1%unsourced
  • SWE-bench Pro: not reported by the vendor
  • MMLU-Pro: not reported by the vendor
  • Terminal-Bench 2.x: benchmark did not exist at release
  • MMMU-Pro: benchmark did not exist at release

Links

More from DeepSeek

Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.