DeepSeek-V4-Flash
DeepSeek- Arena score
- 1432
- Among open models
- #22of 148
- Among all on Arena
- #93
- Votes
- 51,336
Measured on DeepSeek-V4-Flash, the served version, not a local build. LMArena, CC BY 4.0, published Nov 11, 2026.
Score by category
CategoryScoreOpen rankArena rankVotes
- Chinese1464#29#1012.9K
- Coding1457#25#9115.1K
- Software & IT1453#28#9921.1K
- Multi-turn1449#12#719.3K
- Hard prompts, English1449#23#9016K
- Medicine & health1448#25#833.6K
- Science1447#22#878.3K
- Hard prompts1443#20#8934.3K
- Expert1442#29#975.3K
- English1441#24#9522.9K
- French1439#28#1042K
- Legal & government1439#22#934.1K
- Spanish1435#25#931.6K
- Polish1435#17#761K
- Long prompts1435#21#9023K
- Applied math1432#38#1152.9K
- Math1426#33#1102.7K
- Russian1426#19#905.6K
- Business & finance1425#27#9510.3K
- Excluding ties1424#22#9338.8K
- Instructions1421#21#8717.7K
- Non-English1419#21#9228.5K
- German1417#29#108838
- Writing & literature1410#19#9512.4K
- Creative writing1403#19#948.4K
- Media & sports1400#22#9711K
- Japanese1389#24#96645
- Korean1382#24#98871
| Model | Params | Score |
|---|---|---|
| 291B | 1432 | |
| 304B | — |
About as strong, smaller
Compare
gemma-4-26B-A4B-itGoogle143425.8B

Qwen3.8-27BQwen143727.8B

gemma-4-31B-itGoogle144331.3B

Qwen3-Next-80B-A3B-InstructQwen141781.3B
Discussion
- DeepSeek V4-Flash локально | 288GB VRAM на четырёх GPUServerFlow - серверы и серверные комплектующие · Aug 20, 20264.4K likes
- DeepSeek V4 Flash 0731tosh · Aug 7, 2026796 points
- DeepSeek-V4-Flash Updatednhkng · Jul 31, 2026745 points
- DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysistheanonymousone · Jul 31, 2026594 points
- DeepSeek V4 Flash on a Single AMD MI300Xzhoutong · Aug 4, 2026382 points
Built on it
Ways to run it
- GGUF · llama.cpp, Ollama, LM Studiohuihui-ai/Huihui-DeepSeek-V4-Flash-abliterated-ds4-GGUF65 builds
- MLX · Apple silicon: mlx-lm, LM Studiomlx-community/DeepSeek-V4-Flash26 builds
- GPTQ · vLLM, SGLangcanada-quant/DeepSeek-V4-Flash-W4A16-FP8-MTP2 builds
Most downloaded
- Huihui-DeepSeek-V4-Flash-abliterated-ds4-GGUFhuihui-ai · Quantization460.9K downloads
- deepseek-v4-ggufantirez · Quantization297.3K downloads
- DeepSeek-V4-Flash-NVFP4nvidia · Quantization101.1K downloads
- MemT-1.0-FlashMemTensor · Fine-tune57.5K downloads
- DeepSeek-V4-Flash-FP8sgl-project · Quantization23.6K downloads
BenchClub runs
Not runnable yet
Runs open once a DeepSeek-V4-Flash build is registered.
More from DeepSeek
See more
DeepSeek-V3.29.6M685B

DeepSeek-V4-Flash-07312M304B

DeepSeek-V4.1-Flash1.6M763B

DeepSeek-V4-Pro309.2K1.6T
Other 80–300B models
See more
MiniMax-M2.5MiniMax1M229B
Inkling-SmallThinking Machines1M266B
Hy3Tencent987.2K299B

NVIDIA-Nemotron-3-Super-120B-A12B-BF16NVIDIA874.7K124B