DeepSeek-V3.1-Terminus

Home/Models/deepseek-ai/DeepSeek-V3.1-Terminus

deepseek-ai/DeepSeek-V3.1-Terminus

kind: fp8
quant: fp8 (8-bit)
family: deepseek
modality: text-generation
root model: deepseek-ai/DeepSeek-V3.1-Base
provider: deepseek-ai
license: mit
language: —
context: —
library: transformers
vRAM (GB): —
size on disk: —
downloads: 14,951
likes: 366
created: —
last modified: 2025-09-29T01:52:07+00:00 (2025)
huggingface: https://huggingface.co/deepseek-ai/DeepSeek-V3.1-Terminus
tags: transformerssafetensorsdeepseek_v3text-generationconversationalcustom_codearxiv:2412.19437base_model:deepseek-ai/DeepSeek-V3.1-Basebase_model:quantized:deepseek-ai/DeepSeek-V3.1-Baselicense:miteval-resultstext-generation-inferenceendpoints_compatiblefp8region:us

benchmark	category	score	context	conf	evidence
SimpleQA	knowledge	93.4	SimpleQA	medium	`\| SimpleQA \| 93.4 \| 96.8 \|`
MMLU-Pro	knowledge	84.8	MMLU-Pro	medium	`\| MMLU-Pro \| 84.8 \| 85.0 \|`
GPQA Diamond	reasoning	80.1	GPQA-Diamond	medium	`\| GPQA-Diamond \| 80.1 \| 80.7 \|`
LiveCodeBench	coding	74.8	LiveCodeBench	medium	`\| LiveCodeBench \| 74.8 \| 74.9 \|`
BrowseComp	agentic	49.2	BrowseComp-zh	medium	`\| BrowseComp-zh \| 49.2 \| 45.0 \|`
Terminal-Bench	agentic-coding	31.3	Terminal-bench	medium	`\| Terminal-bench \| 31.3 \| 36.7 \|`
BrowseComp	agentic	30.0	BrowseComp	medium	`\| BrowseComp \| 30.0 \| 38.5 \|`
Humanity's Last Exam	knowledge	15.9	Humanity's Last Exam	medium	`\| Humanity's Last Exam \| 15.9 \| 21.7 \|`

none scraped.