DeepSeek cloud models

31 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 (Reasoning, Max Effort)	51.5	76.576
DeepSeek V4 Pro (Reasoning, High Effort)	49.8	74.214
DeepSeek V4 Flash (Reasoning, Max Effort)	46.5	96.704
DeepSeek V4 Flash (Reasoning, High Effort)	46	0
DeepSeek V3.2 (Reasoning)	41.7	0
DeepSeek V4 Pro (Non-reasoning)	39.3	68.599
DeepSeek V4 Flash (Non-reasoning)	36.5	89.443
DeepSeek V3.1 Terminus (Reasoning)	33.9	0
DeepSeek V3.2 Exp (Reasoning)	32.9	0
DeepSeek V3.2 (Non-reasoning)	32.1	0
DeepSeek V3.2 Speciale	29.4	0
DeepSeek V3.1 Terminus (Non-reasoning)	28.5	0
DeepSeek V3.2 Exp (Non-reasoning)	28.4	0
DeepSeek V3.1 (Non-reasoning)	28.1	0
DeepSeek V3.1 (Reasoning)	27.7	0
DeepSeek R1 0528 (May '25)	27.1	0
DeepSeek V3 0324	22.3	0
DeepSeek R1 (Jan '25)	18.8	0
DeepSeek R1 Distill Qwen 32B	17.2	0
DeepSeek V3 (Dec '24)	16.5	0
DeepSeek R1 0528 Qwen3 8B	16.4	0
DeepSeek R1 Distill Llama 70B	16	44.425
DeepSeek R1 Distill Qwen 14B	15.8	0
DeepSeek-V2.5 (Dec '24)	12.5	0
DeepSeek-V2.5	12.3	0
DeepSeek R1 Distill Llama 8B	12.1	0
DeepSeek-Coder-V2	10.6	0
DeepSeek R1 Distill Qwen 1.5B	9.1	0
DeepSeek-V2-Chat	9.1	0
DeepSeek Coder V2 Lite Instruct	8.5	0
DeepSeek LLM 67B Chat (V1)	8.4	0

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