DeepSeek V3.1

+ contribute

DeepSeek V3.1 is an open-weights language model from DeepSeek, released August 2025. Its published size is 685B parameters (37B active per token) and a 131k-token context window. Licence: mit.

65.8±0.6#133 of 719 LLMsfrom 66 benchmark results across 9 sources

$ how it compares

The LLMs ranked either side of DeepSeek V3.1, out of 828 with a score, then more from DeepSeek.

#	model	score	where	RAM	released
150	MiniMax-M2.5 MiniMax	66±1	api $0.53/M +local	~75GB	2026-02
151	Qwen3-Max (2025-09-23) Alibaba	66±1	api $2.40/M		2025-09
152	DeepSeek V3.1 DeepSeek	66±1	api $0.42/M +local	~110GB	2025-08
153	Spark-X2.5 4B iFLYTEK (SparkLLM)	66±1	local	~3.9GB	2026-08
154	MiniMax-M2.1 MiniMax	66±1	local	~79GB	2025-12
	more from DeepSeek
21	DeepSeek V4.1 Flash DeepSeek	78±1	api $0.26/M +local	~537GB	2026-09
31	DeepSeek V4 Pro (0813) DeepSeek	76±1	api $1.63/M +local	~1010GB	2026-08
35	DeepSeek V3.2 Speciale DeepSeek	76±3	comparison only

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #143 of 300 · 21% of its score · 11 tests

Beats Qwen3 235B-A22B Thinking 2507, o4-mini, o1 (2024-12-17) here, though they rank higher overall. Trails SmolLM3 3B, Jamba 1.5 Mini, Jamba 1.5 Large here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
τ²-bench Retail +1	64.9% ✎	▅···█·▃····▃ beats 1/6	medium · avg 36%	▇▇ 4.0%	at its level · several runs
SWE-bench Verified lab +1	66.0% ✎	▁▁▁▁▁▁▂▄█▆▁▁ beats 65/131	hard · avg 24%	▇ 3.5%	sharp · at its level · small run · several runs
τ²-bench Telecom +1	38.5% ✎	██··███···██ beats 3/6	very hard · avg 9.6%	▇ 3.0%	sharp · at its level · noisy · several runs
BrowseComp-ZH lab +1	49.2% ✎	▁▁▁▁▁▃·▁▄▅▄█ beats 12/23	hard · avg 16%	▇ 2.6%	at its level · small run · several runs
SWE-bench Multilingual lab +1	54.5% ✎	▁··▁▃▂▅▇█▃▃▁ beats 14/60	hard · avg 24%	▇ 2.2%	at its level · small run · several runs
τ²-bench Airline +1	46.0% ✎	▄·▄··▄···█▄▄ beats 2/6	hard · avg 29%	▇ 2.2%	at its level · several runs
BrowseComp lab +1	30.0% ✎	▂·▄▁▂▃▆▄█▂▃▄ beats 10/42	very hard · avg 14%	▇ 1.8%	at its level · small run · several runs
Terminal-Bench (pre-2.0 release) lab	31.3% ✎	▃·▃···▃▃▅·▅█ beats 5/10	very hard · avg 14%	▇ 0.8%	at its level · small run

3 more tests

test	result	vs the others in its run	difficulty	counts	why
Humanity's Last Exam with tools lab	29.8% ✎	▂▂▂▁▄▂▃▆█▇▂▆ beats 8/54	very hard · avg 13%	▇ 0.8%	at its level · small run
Terminal-Bench (pre-2.0) with Terminus 1 lab	31.3% ✎	█····█·····█ beats 2/2	very hard · avg 7.1%	▇ 0.3%	noisy · small run
BFCL v3	55.4% ✎	▁▁▁▁▁▃▅█▆▅▄▁ beats 69/170	medium · avg 56%	▇ 0.3%	blunt · at its level

coding▁▁▁▃▆██▇▆▄▃▁ #109 of 458 · 22% of its score · 7 tests

Beats Grok 4.3, Qwen3.5 Plus, DeepSeek V3.2 here, though they rank higher overall. Trails Llama 3.1 Nemotron Ultra 253B v1, Solar Pro 4, Laguna XS 2.1 (33B-A3B) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Aider polyglot (225 exercises, two attempts) lab +1	76.3% ✎	▂▁▁▃▆▃▄▆█▃▄▂ beats 64/71	very hard · avg 14%	▇▇▇▇ 10%	sharp · small run · several runs
SWE-bench Verified lab +1	66.0% ✎	▁▁▁▁▁▁▂▄█▆▁▁ beats 65/131	hard · avg 24%	▇ 3.5%	sharp · at its level · small run · several runs
LiveCodeBench (version unspecified) +1	73.5% ✎	▄······▄█▄·▄ beats 1/5	medium · avg 39%	▇ 3.0%	at its level · several runs
SWE-bench Multilingual lab +1	54.5% ✎	▁··▁▃▂▅▇█▃▃▁ beats 14/60	hard · avg 24%	▇ 2.2%	at its level · small run · several runs
MBPP+	79.9% ✎	▃·····▃█▃··▃ beats 4/6	medium · avg 61%	▇ 1.5%	typical
HumanEval+	92.7% ✎	▄▄▄·····▄▄·█ beats 3/6	easy · avg 67%	▇ 1.4%	sharp · maxed out
LiveCodeBench 2024-08–2025-05 (version unstated) lab	74.8% ✎	▃···█··▅··▃▃ beats 5/7	medium · avg 40%	▇ 0.7%	at its level · small run

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #95 of 393 · 5.1% of its score · 3 tests

Beats LongCat 2.0 (1.6T-A48B), Inkling Small (266B-A12B), GPT-5.4 nano here, though they rank higher overall. Trails Gemma 4 12B, Kimi K2 Instruct (0905), Gemma 4 26B-A4B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
LMArena (text)	1415.9 ✎	▁▁▁▂▃▃▃▆▆▇█▃ beats 265/379	avg 1268	▇ 2.9%	blunt
IFEval (aggregation unspecified)	86.7% ✎	█··█·█··████ beats 2/6	easy · avg 75%	▇ 1.2%	typical
IFEval prompt-level strict	86.3% ✎	███··█·█···█ beats 1/5	easy · avg 74%	▇ 1.0%	typical

math▁▂▃▅▆▅▄█▄▁▁▁ #92 of 398 · 16% of its score · 4 tests

Beats Claude Sonnet 4.6, Claude Opus 4.5, MiniMax-M3 here, though they rank higher overall. Trails Gemini 2.5 Flash, Granite 4.2 30B, Qwen3 235B-A22B Instruct 2507 here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
AIME 2025 +2	49.3% ✎	▄▄···▄▄··▄·█ beats 2/6	medium · avg 51%	▇▇▇ 6.3%	sharp · several runs
AIME 2024 +1	93.9% ✎	▁▁▁▂▁▃▁▅███▃ beats 84/85	medium · avg 60%	▇▇ 4.3%	sharp · several runs
HMMT 2025 (edition unspecified) +1	80.4% ✎	▄··▄··▄·▄··█ beats 3/5	hard · avg 29%	▇▇ 4.0%	sharp · several runs
MATH-500 +1	96.1% ✎	▄·▄··█··▄·▄▄ beats 2/6	very easy · avg 91%	▇ 1.5%	maxed out · noisy · several runs

web search█········█·█ #2 of 3 · 5.2% of its score · 3 tests

test	result	vs the others in its run	difficulty	counts	why
BrowseComp-ZH lab +1	49.2% ✎	▁▁▁▁▁▃·▁▄▅▄█ beats 12/23	hard · avg 16%	▇ 2.6%	at its level · small run · several runs
BrowseComp lab +1	30.0% ✎	▂·▄▁▂▃▆▄█▂▃▄ beats 10/42	very hard · avg 14%	▇ 1.8%	at its level · small run · several runs
Humanity's Last Exam with tools lab	29.8% ✎	▂▂▂▁▄▂▃▆█▇▂▆ beats 8/54	very hard · avg 13%	▇ 0.8%	at its level · small run

other tests19 not in a task · 41% of its score

test	result	vs the others in its run	difficulty	counts	why
GPQA Diamond +2	84.2% ✎	▄··▄··▄▄···█ beats 5/5	hard · avg 43%	▇▇▇▇ 10%	sharp · at its level · several runs
MMLU-Pro +3	84.5% ✎	█▄·▄···▄··▄▄ beats 5/6	medium · avg 59%	▇▇▇ 7.0%	several runs
MMLU-Redux +1	90.5% ✎	████··█····█ beats 3/5	medium · avg 73%	▇ 3.3%	several runs
MMLU	91.0% ✎	▄·····█▄▄▄·▄ beats 5/6	medium · avg 71%	▇ 3.0%	typical
VitaBench +1	20.3% ✎	█·······▄█·█ beats 4/6	very hard · avg 7.8%	▇ 2.7%	several runs
Humanity's Last Exam lab +1	15.9% ✎	▁▁▁▃▆▆▇██▆▁▃ beats 37/104	very hard · avg 4.6%	▇ 2.5%	small run · several runs
ZebraLogic +1	85.3% ✎	█···█··███·█ beats 2/5	easy · avg 75%	▇ 2.1%	maxed out · several runs
CMMLU	88.0% ✎	█····▄··▄▄▄▄ beats 4/6	medium · avg 67%	▇ 1.5%	typical

11 more tests

test	result	vs the others in its run	difficulty	counts	why
C-Eval	89.2% ✎	█····▄·▄·▄▄▄ beats 3/6	medium · avg 68%	▇ 1.4%	typical
OJBench	33.6% ✎	▄······▄▄·▄█ beats 2/5	very hard · avg 12%	▇ 1.3%	at its level
BeyondAIME +1	36.5% ✎	█····█·▄▄··▄ beats 2/6	medium · avg 41%	▇ 1.3%	at its level · several runs
COLLIE (constrained text generation)	43.8% ✎	▄···▄█▄····█ beats 0/6	hard · avg 30%	▇ 1.2%	at its level
RPC-Bench overall, text input (Z.ai)	30.8% ✎	▄··▄▄·▄▄█·▄▄ beats 4/8	very hard · avg 11%	▇ 1.2%	at its level
CPTUB average (SpeakLeash, 0-5 judge scale)	80.5% ✎	▁▂▂▂▃█▃▆▅▄▂▃ beats 83/95	easy · avg 66%	▇ 1.2%	typical
DROP	84.2% ✎	▄···▄·▄█·▄·▄ beats 5/6	medium · avg 41%	▇ 1.1%	noisy
Open PT-LLM Leaderboard average (9 Portuguese tasks)	82.6% ✎	▃▁▁▁▁▁▁▁▂█▅▃ beats 270/276	medium · avg 73%	▇ 0.4%	noisy
miniF2F (pass@1)	49.6% ✎	██······▄··▄ beats 4/5	very hard · avg 13%	▇ 0.1%	blunt · at its level · noisy
Vectara hallucination leaderboard, 2025 dataset (HHEM-2.3, factual consistency rate)	94.5% ✎	▁▁▂▃▄▄█▆▃▂▂▁ beats 54/140	very easy · avg 94%	▇ 0.1%	blunt · maxed out · noisy
Vectara hallucination leaderboard (HHEM-2.3, factual consistency rate)	94.5% ✎	▁▂▁▂▅█▂▃▄▃▁▁ beats 82/98	very easy · avg 90%	▇ 0.0%	blunt

$ where to run it

Through an API

api · Deep Infra · $0.42/M · $0.25 in / $0.95 out per 1M tokens · 131k context · deepinfra.com ↗ · price sourceStandard base token rate from models.dev. Provider-specific tiers, long-context surcharges, caching and promotions may differ.

llama.cpp · recommended

version

Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
macOS / Linux
curl -LsSf https://llama.app/install.sh | sh
Windows
irm https://llama.app/install.ps1 | iex
other ways to install ↗
Run DeepSeek V3.1 · needs ~411.5 GB · from huggingface.co/unsloth
chat in the terminal
llama cli -hf unsloth/DeepSeek-V3.1-GGUF:UD-Q4_K_XL
Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf unsloth/DeepSeek-V3.1-GGUF:UD-Q4_K_XL

LM Studio

version

Install LM Studio · Open the app once before using the lms command.
download it (macOS / Windows / Linux) ↗
Run DeepSeek V3.1 · needs ~411.5 GB · from huggingface.co/unsloth
open in LM Studio ↗
Or download it in the terminal
lms get https://huggingface.co/unsloth/DeepSeek-V3.1-GGUF@UD-Q4_K_XL -y
Then run lms ls to see its name, and lms chat followed by that name

MLX (Mac)

version

Install MLX (Mac) · Apple silicon Macs only, text only.
with uv
uv tool install mlx-lm
or
pip install mlx-lm
Run DeepSeek V3.1 · needs ~377.6 GB, Mac only · from huggingface.co/mlx-community
chat in the terminal
mlx_lm.chat --model mlx-community/DeepSeek-V3.1-4bit
Or start an OpenAI-compatible server at http://localhost:8080/v1
mlx_lm.server --model mlx-community/DeepSeek-V3.1-4bit

Weights and documentation: huggingface.co/deepseek-ai ↗

$ how the score was worked out · Meituan LongCat-Flash-Thinkin… 34%, Meituan LongCat-Flash-Chat mo… 23%, deepseek-ai/DeepSeek-V3.1 mod… 21% and 6 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

source	results	share of the score	licence
Meituan LongCat-Flash-Thinking model card (evaluation table)	18	▇▇▇▇▇▇▇▇▇▇ 34%	MIT
Meituan LongCat-Flash-Chat model card (evaluation table)	17	▇▇▇▇▇▇▇ 23%	MIT
deepseek-ai/DeepSeek-V3.1 model card · lab-reported	15	▇▇▇▇▇▇ 21%	MIT
DeepSeek-V3.1-Terminus model card · lab-reported	10	▇▇▇▇▇ 16%	MIT
LMArena leaderboard dataset	1	▇ 3%	CC BY 4.0
RPC-Bench leaderboard (Z.ai research paper comprehension)	1	▇ 1%	Apache-2.0
SpeakLeash CPTUB (Complex Polish Text Understanding Benchmark)	1	▇ 1%	MIT
Open Portuguese LLM Leaderboard (archived)	1	▇ 0%	Apache-2.0
Vectara hallucination leaderboard (vectara/hallucination-leaderboard repository)	2	▇ 0%	Apache-2.0
prior · pull toward the middle		▇ 0%
weighted average		= 65.8

$ improve this page

Update model information Add a place to run it Add a test result Propose a benchmark Correct a benchmark score Edit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated 2026-10-11 02:20 UTC · sources and licences · json