DeepSeek-V3
DeepSeek- Arena score
- 1333
- Among open models
- #77of 148
- Among all on Arena
- #229
- Votes
- 21,770
Measured on DeepSeek-V3, the served version, not a local build. LMArena, CC BY 4.0, published Nov 11, 2026.
Score by category
CategoryScoreOpen rankArena rankVotes
- Spanish1352#61#193127
- Multi-turn1349#63#2093.7K
- English1346#81#23312.9K
- Medicine & health1345#70#2191.1K
- Long prompts1343#65#2103.1K
- French1343#61#197242
- Legal & government1341#70#2201.3K
- Chinese1338#76#2291.5K
- Software & IT1336#82#2395.2K
- Science1333#78#2323.8K
- Writing & literature1331#62#2045.9K
- Creative writing1329#61#1973.6K
- Coding1325#81#2423.3K
- Russian1324#70#2172.5K
- Media & sports1324#60#2014.2K
- German1323#60#193493
- Hard prompts, English1323#84#2473.6K
- Non-English1316#73#2248.9K
- Instructions1315#72#2248.6K
- Applied math1314#79#2342.3K
- Hard prompts1312#81#2435.4K
- Math1311#78#2332.7K
- Business & finance1307#83#2402.2K
- Expert1305#80#2341.2K
- Excluding ties1289#75#22614.1K
- Japanese1267#51#182534
- Korean1249#62#204295
| Model | Params | Score |
|---|---|---|
| 684B | 1333 | |
| 685B | 1375 |
About as strong, smaller
Compare
granite-4.2-8bIBM Granite13198.8B
gemma-3-12b-itGoogle133412.2B

Mistral-Small-3.2-24B-Instruct-2506Mistral AI133824.0B

gemma-4-26B-A4B-itGoogle143425.8B
Discussion
1 post this week
Built on it
Ways to run it
- GGUF · llama.cpp, Ollama, LM Studiounsloth/DeepSeek-V3-GGUF4 builds
- MLX · Apple silicon: mlx-lm, LM Studiomlx-community/DeepSeek-V3-4bit4 builds
- AWQ · vLLM, SGLangQuixiAI/DeepSeek-V3-AWQ3 builds
- GPTQ · vLLM, SGLangOPEA/DeepSeek-V3-int4-sym-gptq-inc2 builds
Most downloaded
BenchClub runs
Not runnable yet
Runs open once a DeepSeek-V3 build is registered.
More from DeepSeek
See more
DeepSeek-V3.29.6M685B

DeepSeek-V4-Flash-07312M304B

DeepSeek-V4.1-Flash1.6M763B

DeepSeek-V4-Flash1.2M291B
Other >300B models
See more
GLM-5.3-FlashZ.ai6.4M321B
InklingThinking Machines1.5M952B
NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16NVIDIA1M561B

Kimi-K3Moonshot AI768.9K2.8T