DE
DeepSeek
Language model · China ·Sep 2025

DeepSeek-V3.1-Terminus

Language Open (Restricted)
671B
Parameters
50
Benchmarks

About DeepSeek-V3.1-Terminus

DeepSeek-V3.1-Terminus is an AI model developed by DeepSeek, in the language category, released in 2025, made available as an open-weights model with 671B params.

On this page you'll find DeepSeek-V3.1-Terminus's full specifications, including 50 benchmark results. Review provider pricing and benchmark scores below, or compare DeepSeek-V3.1-Terminus head-to-head with other language models.

Links

Run DeepSeek-V3.1-Terminus locally — what it takes

DeepSeek-V3.1-Terminus (671B) is too large to run on a single consumer or Apple machine. At Q4 its weights alone are about 376 GB — it needs a multi-GPU / datacenter setup:

  • ≈ 2× AMD Instinct MI325X
  • ≈ 3× NVIDIA B200

Lower quantization (Q3/Q2) or CPU + system-RAM offload (e.g. ktransformers) reduces the requirement at the cost of speed.

Software support

✓ measured · · compatible · — not supported. Informational only — speed is hardware-based.

Community

Would you run DeepSeek-V3.1-Terminus again?
No votes yet — be the first.
Did it run? — community reports by hardware

No reports yet. Own this model on some hardware? Be the first to confirm it runs.

Discussion

No comments yet. Start the discussion.

DeepSeek-V3.1-Terminus benchmark scores
Benchmark Category Score Variant
Codeforces Elo code 2,073 thinking-rating-cited-ring-1t
Codeforces Elo code 1,582 non-thinking-rating-cited-ling-1t
FullStackBench code 55.48% non-thinking-cited-ling-1t
LiveCodeBench code 48.02% non-thinking-cited-ling-1t
LiveCodeBench v6 code 75.33% thinking-cited-ring-1t
MBPP code 90.69% sanitized-non-thinking-cited-ling-1t
Aider Benchmark coding 92.86 thinking-cited-ring-1t
Aider Benchmark coding 88.16 non-thinking-edit-cited-ling-1t
MultiPL-E coding 77.68 non-thinking-cited-ling-1t
ArenaHard v2 general 63.24 non-thinking-win-rate-cited-ling-1t
ArenaHard v2 general 60.27 thinking-win-rate-cited-ring-1t
C-Eval general 91.76% non-thinking-cited-ling-1t
C-Eval general 91.22% thinking-cited-ring-1t
IFEval general 89.09% thinking-cited-ring-1t
IFEval general 86.32% non-thinking-prompt-strict-cited-ling-1t
MMLU-Pro general 85% thinking-cited-ring-1t
MMLU-Pro general 83.25% non-thinking-cited-ling-1t
MMLU-Redux general 92.37% non-thinking-cited-ling-1t
MultiChallenge general 45.79% thinking-cited-ring-1t
MultiChallenge general 42.49% non-thinking-cited-ling-1t
AGIEval knowledge 89.83 thinking-cited-ring-1t
CMMLU knowledge 89.2 thinking-cited-ring-1t
TriviaQA knowledge 82.77 thinking-cited-ring-1t
AIME 2024 math 71.67% non-thinking-cited-ling-1t
AIME 2025 math 89.06% thinking-cited-ring-1t
AIME 2025 math 55.21% non-thinking-cited-ling-1t
CNMO 2024 math 85.42% thinking-cited-ring-1t
CNMO 2024 math 73.78% non-thinking-cited-ling-1t
HMMT February 2025 math 86.1% thinking-cited-ring-1t
HMMT February 2025 math 41.25% non-thinking-cited-ling-1t
Omni-MATH math 81.93 thinking-cited-ring-1t
Omni-MATH math 64.77 non-thinking-cited-ling-1t
UGMathBench math 77.19 thinking-cited-ring-1t
UGMathBench math 72.7 non-thinking-cited-ling-1t
ARC-AGI-1 reasoning 40.62 thinking-cited-ring-1t
ARC-AGI-1 reasoning 14.69 non-thinking-cited-ling-1t
BBEH reasoning 61.04 thinking-cited-ring-1t
BBEH reasoning 42.86 non-thinking-cited-ling-1t
FinanceReasoning reasoning 87.76 thinking-cited-ring-1t
FinanceReasoning reasoning 86.44 non-thinking-cited-ling-1t
GPQA Diamond reasoning 81% thinking-cited-ring-1t
GPQA Diamond reasoning 76.23% non-thinking-cited-ling-1t
Humanity's Last Exam reasoning 17.82% thinking-cited-ring-1t
Humanity's Last Exam reasoning 10.38% non-thinking-cited-ling-1t
ZebraLogic reasoning 96.33% thinking-cited-ring-1t
ZebraLogic reasoning 81.6% non-thinking-cited-ling-1t
HealthBench safety 50.19% thinking-cited-ring-1t
PhyBench science 47.91 thinking-cited-ring-1t
BFCL v3 tool-use 62.01% thinking-cited-ring-1t
BFCL v3 tool-use 52.67% non-thinking-cited-ling-1t

More models like DeepSeek-V3.1-Terminus

Frequently asked questions

What is DeepSeek-V3.1-Terminus?

DeepSeek-V3.1-Terminus is an AI model developed by DeepSeek, in the language category, released in 2025. It is tracked on GenAIList with its specifications, benchmark scores and provider pricing.

Who created DeepSeek-V3.1-Terminus?

DeepSeek-V3.1-Terminus was developed by DeepSeek and released in 2025.

Is DeepSeek-V3.1-Terminus open source or proprietary?

DeepSeek-V3.1-Terminus ships as an open-weights model: you can download and self-host the weights, though the licence may place some restrictions on use.

How much does DeepSeek-V3.1-Terminus cost?

Pricing for DeepSeek-V3.1-Terminus depends on the provider. See the providers table on this page for the latest API rates.

How does DeepSeek-V3.1-Terminus perform on benchmarks?

DeepSeek-V3.1-Terminus is benchmarked across 50 evaluations on GenAIList, including Codeforces Elo (2,073). See the full benchmark table below and compare it with other models.

No videos yet. Know a good one? Add it below.

Know a great video about DeepSeek-V3.1-Terminus?