DGX Spark, GB10 Grace Blackwell, 128GB

4TB of storage, 273GB/s of memory bandwidth, and 120GB of unified LPDDR5X memory left for a model once DGX OS has taken its share. It holds 21 of the 30 models we track.

Six other boxes ship the same GB10 chip and the same 128GB at the same 273GB/s. One of them, the ASUS Ascent GX10, is on the Grid in its own right; ASUS raised its own price above this one in September, so check both.

Also sold on the same chip, same 128GB, same 273GB/s, all dearer than the two on the Grid, priced 28 Aug 2026:

HP ZGX Nano G1n — $5,199, storage tier unconfirmed
Dell Pro Max with GB10 — $5,654.39, 4TB
Lenovo ThinkStation PGX — $5,849, 1TB
Acer Veriton GN100 — $6,691.49, 4TB

Offers

checked 46 days ago

NVIDIAlist pricecheapest todayin stockchecked 46 days ago$4,699
Best Buyin stockchecked 46 days ago$5,299.99
$601 over NVIDIA's own price for the same box.

No reseller undercuts NVIDIA on this Config, so NVIDIA direct is the price to beat.

Timing

Buy now. $4,699 is the lowest price in 47 days.

27 Aug 2026$4,699
28 Aug 2026$5,299.99

Models it runs — 8K context

Llama-4-Behemoth—
Won't fit— headroom
DeepSeek-V4-Pro—
Won't fit— headroom
Kimi-K2-Instruct—
Won't fit— headroom
DeepSeek-V3—
Won't fit— headroom
DeepSeek-R1—
Won't fit— headroom
Qwen3-Coder-480B-A35B-Instruct—
Won't fit— headroom
Llama-4-Maverick-17B-128E—
Won't fit— headroom
GLM-4.6—
Won't fit— headroom
DeepSeek-V4-Flash—
Won't fit— headroom
Qwen3-235B-A22BQ3_K_M
Tight3GB headroomest. 12.7 tok/s
MiniMax-M2Q3_K_M
Tight5GB headroomest. 27.9 tok/s
Mistral-Large-2411Q6_K
Fits15GB headroomest. 1.8 tok/s
gpt-oss-120bQ6_K
Fits23GB headroomest. 32.6 tok/s
Llama-4-Scout-17B-16EQ8_0
Tight2GB headroomest. 7.6 tok/s
Qwen3-Coder-NextQ8_0
Fits34GB headroomest. 42.8 tok/s
Llama-3.3-70BQ8_0
Fits42GB headroomest. 2.4 tok/s
Qwen3.5-35B-A3BBF16
Fits49GB headroomest. 22.8 tok/s
Qwen3-32BBF16
Fits53GB headroomest. 2.8 tok/s
Gemma-4-31BBF16
Fits56GB headroomest. 2.9 tok/s
Qwen3-30B-A3BBF16
Fits57GB headroomest. 20.7 tok/s
Gemma-3-27BBF16
Fits63GB headroomest. 3.3 tok/s
Qwen3.6-27BBF16
Fits65GB headroomest. 3.3 tok/s
Gemma-4-26B-A4BBF16
Fits68GB headroomest. 18.0 tok/s
Devstral-Small-2507BF16
Fits70GB headroomest. 3.7 tok/s
gpt-oss-20bBF16
Fits77GB headroomest. 19.0 tok/s
Phi-4BF16
Fits89GB headroomest. 6.3 tok/s
Gemma-4-12BBF16
Fits95GB headroomest. 7.4 tok/s
Qwen3-VL-8B-InstructBF16
Fits102GB headroomest. 11.1 tok/s
Gemma-3n-E4BBF16
Fits105GB headroomest. 22.2 tok/s
Qwen3-4BBF16
Fits110GB headroomest. 22.2 tok/s

Model	Best quant	Fit	Headroom	Est. speed
Llama-4-Behemoth	—	Won't fit	—	—
DeepSeek-V4-Pro	—	Won't fit	—	—
Kimi-K2-Instruct	—	Won't fit	—	—
DeepSeek-V3	—	Won't fit	—	—
DeepSeek-R1	—	Won't fit	—	—
Qwen3-Coder-480B-A35B-Instruct	—	Won't fit	—	—
Llama-4-Maverick-17B-128E	—	Won't fit	—	—
GLM-4.6	—	Won't fit	—	—
DeepSeek-V4-Flash	—	Won't fit	—	—
Qwen3-235B-A22B	Q3_K_M	Tight	3GB	12.7 tok/s
MiniMax-M2	Q3_K_M	Tight	5GB	27.9 tok/s
Mistral-Large-2411	Q6_K	Fits	15GB	1.8 tok/s
gpt-oss-120b	Q6_K	Fits	23GB	32.6 tok/s
Llama-4-Scout-17B-16E	Q8_0	Tight	2GB	7.6 tok/s
Qwen3-Coder-Next	Q8_0	Fits	34GB	42.8 tok/s
Llama-3.3-70B	Q8_0	Fits	42GB	2.4 tok/s
Qwen3.5-35B-A3B	BF16	Fits	49GB	22.8 tok/s
Qwen3-32B	BF16	Fits	53GB	2.8 tok/s
Gemma-4-31B	BF16	Fits	56GB	2.9 tok/s
Qwen3-30B-A3B	BF16	Fits	57GB	20.7 tok/s
Gemma-3-27B	BF16	Fits	63GB	3.3 tok/s
Qwen3.6-27B	BF16	Fits	65GB	3.3 tok/s
Gemma-4-26B-A4B	BF16	Fits	68GB	18.0 tok/s
Devstral-Small-2507	BF16	Fits	70GB	3.7 tok/s
gpt-oss-20b	BF16	Fits	77GB	19.0 tok/s
Phi-4	BF16	Fits	89GB	6.3 tok/s
Gemma-4-12B	BF16	Fits	95GB	7.4 tok/s
Qwen3-VL-8B-Instruct	BF16	Fits	102GB	11.1 tok/s
Gemma-3n-E4B	BF16	Fits	105GB	22.2 tok/s
Qwen3-4B	BF16	Fits	110GB	22.2 tok/s

Best quant is the highest-quality one that still loads with headroom. Every speed is an estimate; the formula is here.