DE
DeepSeek
Language model · China ·Aug 2025

DeepSeek-V3.1

Language Open (Restricted)
671B
Parameters
14
Benchmarks

About DeepSeek-V3.1

DeepSeek-V3.1 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's full specifications, including 14 benchmark results. Review provider pricing and benchmark scores below, or compare DeepSeek-V3.1 head-to-head with other language models.

→ More models by DeepSeek → All language models → AI benchmarks leaderboard

Links

Announcement ↗ Hugging Face ↗

Run DeepSeek-V3.1 locally — what it takes

DeepSeek-V3.1 (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

·ExLlamaV2 ·KoboldCpp ·LM Studio ·MLX ·Ollama ·SGLang ·TensorRT-LLM ·llama.cpp ·vLLM ?KTransformers ?text-generation-webui

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

Community

Would you run DeepSeek-V3.1 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 benchmark scores▶
Compare on the benchmark table →
Benchmark Category Score Variant
BrowseComp agentic 30% react-cited-tongyi
BrowseComp-Zh agentic 49.2 react-cited-tongyi
GAIA agentic 63.1 react-cited-tongyi
WebWalkerQA agentic 61.2 react-cited-tongyi
xbench-DeepSearch agentic 71 react-cited-tongyi
LiveCodeBench v5 code 66.59%
Frames long-context 83.7 react-cited-tongyi
AIME 2024 math 91.87%
AIME 2025 math 82.49%
HMMT February 2025 math 83.54%
Omni-MATH-HARD math 53.22%
GPQA Diamond reasoning 79.46%
Humanity's Last Exam reasoning 29.8% react-cited-tongyi
Humanity's Last Exam reasoning 8.4%

More models like DeepSeek-V3.1

AI
DMind-1
 AI
COLLECTIVE-1
 AI
Cogito 1 series
 AI
Euler (Euler大模型)
· 2.5B
 AI
SmolLM-1.7B
· 1.7B
 AI
Llama-3-70B-Synthia-v3.5
· 70B

Frequently asked questions

What is DeepSeek-V3.1? ▶

DeepSeek-V3.1 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? ▶

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

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

DeepSeek-V3.1 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 cost? ▶

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

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

DeepSeek-V3.1 is benchmarked across 14 evaluations on GenAIList, including BrowseComp (30%). 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?