DeepSeek cloud models

34 models tracked via Artificial Analysis. Compare cloud performance, then find local GGUF versions in the GraySoft model catalog.

Model	Intelligence	Speed (tok/s)
DeepSeek V4 Pro 0813 (Reasoning, Max Effort)	42.1	59.459
DeepSeek V4 Flash Vision (Reasoning, Max Effort)	41.5	121.308
DeepSeek V4 Flash 0731 (Reasoning, Max Effort)	40.8	118.418
DeepSeek V4 Pro (Reasoning, Max Effort)	35.8	61.264
DeepSeek V4 Pro (Reasoning, High Effort)	34.6	62.907
DeepSeek V4 Flash (Reasoning, Max Effort)	33.1	0
DeepSeek V4 Flash (Reasoning, High Effort)	30.4	0
DeepSeek V3.2 (Reasoning)	25.3	0
DeepSeek V4 Pro (Non-reasoning)	24.5	59.136
DeepSeek V3.1 Terminus (Reasoning)	23.5	0
DeepSeek V4 Flash (Non-reasoning)	22.1	0
DeepSeek V3.2 Exp (Reasoning)	19	0
DeepSeek V3.2 (Non-reasoning)	18.3	0
DeepSeek V3.2 Speciale	16	0
DeepSeek V3.1 Terminus (Non-reasoning)	15.2	0
DeepSeek V3.2 Exp (Non-reasoning)	15.1	0
DeepSeek V3.1 (Non-reasoning)	14.8	0
DeepSeek V3.1 (Reasoning)	14.5	0
DeepSeek R1 0528 (May '25)	13.9	0
DeepSeek R1 (Jan '25)	12	0
DeepSeek V3 0324	9.2	0
DeepSeek V3 (Dec '24)	8.3	0
DeepSeek R1 Distill Qwen 32B	5.3	0
DeepSeek R1 0528 Qwen3 8B	4.7	0
DeepSeek R1 Distill Llama 70B	4.2	0
DeepSeek R1 Distill Qwen 14B	4.1	0
DeepSeek-V2.5 (Dec '24)	1.2	0
DeepSeek-V2.5	1.1	0
DeepSeek-Coder-V2	1	0
DeepSeek R1 Distill Llama 8B	1	0
DeepSeek LLM 67B Chat (V1)	1	0
DeepSeek R1 Distill Qwen 1.5B	1	0
DeepSeek Coder V2 Lite Instruct	1	0
DeepSeek-V2-Chat	1	0

Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.