Pairs

111 ranked

Took the throneQwen3.8-27Bvs Sep 15Aug · Qwen3.6-35B-A3B · heldSep · Qwen3.8-27BNations

Liberated

#	Model · Build	Hardware			Runtime	Proof
1NEW	Qwen3.8-27B4bit	RTX 5090	370	—	TensorFold 0.6.6	@Oluwaphilemon1
2NEW	Qwen3.5 2BQ4_K_M	RTX 5090	351	—	—	@fillagrew
3NEW	Qwen3.8-27BNVFP4	RTX PRO 6000 Blackwell Max-Q	272.8	—	TensorFold	@llm_frontier_jp
4NEW	GLM 5.3 FlashEXL3	4× RTX PRO 6000	264	—	TensorFold 0.6.2	@Kurcide
5NEW	Qwen3.8-27BEXL3	RTX 5090	258	—	ExLlamaV3	@Oluwaphilemon1
6NEW	GLM 5.3 FlashEXL3	2× RTX PRO 6000	238	—	TensorFold 0.6.5	@Kurcide
7NEW	Swift 1.5 Qwen3.8 Flash-NextIQ2_XS	RTX 4090	220	3830	Strata	@Knuckles_XBT
8NEW	Qwen3.8-Flash-NextIQ3_S	RTX 5090	195.8	3944	Strata	@therealskrlr
9NEW	Qwen3.8-Flash-NextQ2_0	RTX 5090	191	—	Strata	@orenoa7
10NEW	Qwen3.8-27BIQ4_XS	RTX 5090	186	—	llama.cpp	@Oluwaphilemon1
11NEW	Qwen3.8-27BNVFP4	RTX A5000	177.6	—	TensorFold	@Ja6ek
12NEW	Gemma-4 26B-A4B-itUncensoredQ4_K_M	RTX 5090	173	—	—	@fillagrew
13NEW	Qwen3.8-Flash-NextIQ3_S	2x RTX 3090 24GB	158.3	3149	Strata	@geldeki
14NEW	Qwen3.8-Flash-NextIQ3_S	2x RTX 3090 24GB	153.4	2844	Strata	@geldeki
15NEW	Qwen3.8-27BNVFP4	RTX PRO 6000 Blackwell Max-Q	151.4	—	vLLM	@llm_frontier_jp
16NEW	Qwen3.8-Flash-NextIQ3_S	RTX 5090	142	—	Strata	@arianpg
17NEW	Qwen3.8-27BNVFP4	RTX A5000	139.3	—	vLLM	@Ja6ek
18NEW	Qwen3.8-Flash-Next CoderIQ1_M	RTX 5060 Ti 16GB + RTX 3070 8GB	127	—	Strata	@VectorCrossProd
19NEW	GLM 5.3 Flashunknown	DGX Spark	127	—	TensorFold	@Tech2Wild
20NEW	Qwen3.8-Flash-Next	M5 Max	126.5	—	MTPLX V2.11.3	@Youssofal_
21NEW	Qwen3.8-Flash-NextUncensoredIQ3_S	2x RTX 3090 24GB	122.5	2100	Strata 0.1.39	@geldeki
22NEW	GLM 5.3 FlashEXL3 4bpw	RTX PRO 6000 + 2× GB10	120	—	TensorFold	@hokazuya
23NEW	Qwen3.8-Flash-NextUD-Q4_K_XL	RTX 5090	114.3	2141.2	llama.cpp	@mizuirot
24▼23	Qwen3.6-35B-A3B4-bit	M4 Pro 64GB	108.6	—	Rapid-MLX	@draslan_eth
25NEW	GLM 5.3 FlashEXL3 4bpw	DGX Spark	104	—	—	@0xSero
26NEW	Qwen3.8-Flash-NextIQ3_S	RTX 5090	104	1657	Strata	@DaSun64125381
27NEW	DeepSeek v4.1 Flash2.70 bpw	DGX Spark	103	2092	vLLM	@sudoingX
28NEW	Qwen3.8-27B4-bit	DGX Spark	102.9	—	TensorFold	@Oluwaphilemon1
29NEW	Qwen3.8-Flash-NextIQ2_XS	RTX 3090 24GB	100.6	—	Strata	@draslan_eth
30NEW	Qwen3.8-Flash-NextIQ3_S	RTX 3090 24GB	100	—	Strata	@RubenTorney
31NEW	Qwen3.8-Flash-NextIQ3_S	2× RTX 5060 Ti 16GB	99.9	—	Strata	@needmorevram
32▼30	Qwen3.6-35B-A3BUncensoredNVFP4	DGX Spark	94.4	181.5	—	@bonellisystems
33NEW	DeepSeek v4.1 Flashunspecified	DGX Spark	93.6	—	TensorFold	@Tech2Wild
34NEW	Qwen3.8-Flash-NextQ4_XL	RTX PRO 6000	92.8	—	llama.cpp b11412	@piyo621
35NEW	Qwen3.8-Flash-NextUD-IQ4_XS	RTX 4090	92	2850	Strata	@dextune
36NEW	Qwen3.8-27B4-bit	M5 Pro 64GB	89.9	—	TensorFold	@aartiles24
37NEW	Gemma-4 26B-A4B-itUD-Q4_K_XL	Radeon RX 9060 XT 16GB	88	2464	llama.cpp v0.6.0	@RadxaNaoki
38NEW	Qwen3.8-Flash-NextIQ2_XS	RTX 3090 24GB	87	—	Strata	@superalesha
39▼36	Qwen3.6-35B-A3B4bit	M4 Pro	85.5	—	rapid-mlx	@rapidmlx
40NEW	GLM 5.3 FlashEXL3	DGX Spark	85	—	TensorFold	@MehdiHacks
41NEW	Qwen3.5-4B4bit	M4 Pro	82.8	—	rapid-mlx	@rapidmlx
42NEW	GLM 5.3NVFP4	2× DGX Station GB300	81.5	—	SGLang	@SJRonanMD
43NEW	Qwen3.8-Flash-NextNVFP4	RTX 5090	80.9	—	—	@tekizaihq
44NEW	Nemotron 3.5 Bolt 74B2-bit	DGX Spark	80	—	llama.cpp	@filicroval
45NEW	Qwen3.8-Flash-NextEXL3	DGX Spark	79.5	—	EXL3	@yume_arasaki
46NEW	Qwen3.8-Flash-NextIQ2_XS	RTX 3080 Ti 12GB	79	2225	Strata	@nigelinstant
47NEW	Qwen3.8-Flash-NextNVFP4	DGX Spark	74.8	—	TensorFold	@redp314
48NEW	Qwen3.8-27BNVFP4	DGX Spark	72	—	—	@hasso5703
49NEW	Qwen3.8-Flash-NextIQ3_S	RTX 3090 24GB	72	—	Strata	@needmorevram
50NEW	Qwen3.8-Flash-NextIQ3_S	RTX 3090 24GB	70	—	Strata	@superalesha
51NEW	Qwen3.8-Flash-NextIQ3_XXS	RTX 5070 12GB	66.7	—	Strata	@jun1228909
52NEW	AliceAI-Foundation-80B-A3BQ4_K_M	M5 Max	65	—	llama.cpp	@aqty
53NEW	Qwen3.8-Flash-Next4-bit QAD	DGX Spark	65	—	TensorFold	@sudoingX
54NEW	Qwen3.8-Flash-NextFP4	DGX Spark	65	—	NVIDIA	@jtdavies
55NEW	Qwen3.8-27B4-bit	M6 mini 32GB	61.9	—	TensorFold	@jmurillocode
56NEW	GLM 5.3 FlashEXL3	DGX Spark	60.4	1950	TensorFold	@MiaAI_lab
57NEW	Qwen3.8-Flash-NextQ4_K_XL-DN4	4× Tesla V100 32GB	60	—	Strata 0.1.43.1	@DaSun64125381
58NEW	Qwen3.8-Flash-NextFP8	DGX Spark	58	—	vLLM	@sudoingX
59NEW	DeepSeek v4.1 FlashQ2	RTX PRO 6000	57.7	—	llama.cpp b11412	@piyo621
60NEW	Qwen3.8-Flash-NextUD-IQ4_XS	Tesla V100 32GB	57	—	Strata	@tkysyskw
61NEW	GLM 5.3 FlashNVFP4	DGX Spark	55.6	—	—	@PlusTen_AI
62NEW	Qwen3.8-Flash-NextIQ3_S	RTX 4060 Ti	55	—	Strata 0.1.44.1	@xiaobei1943
63NEW	Qwen3.8-Flash-NextIQ3_S	RTX 4060 Ti	55	—	Strata	@xiaobei1943
64NEW	Ternary Bonsai 2 27BPTQ1_0	RTX 5060 Ti	54.7	—	llama.cpp	@niklaslenz_ai
65NEW	CYBER-FROST 3.83.87 bpw	DGX Spark	54.7	—	ExLlamaV3	@niklaslenz_ai
66NEW	Qwen3.8-Flash-NextUD-IQ4_XS	Ryzen AI Max+ 395 128GB	53.8	1293	Strata 0.1.40	@Niko1221
67NEW	Qwen3.8-Flash-NextIQ3_S	RTX 5070 Ti 16GB	53.5	—	Strata	@raminaol9ykas
68NEW	U Underdog Saluki 27B2-bit	DGX Spark	53	—	llama.cpp	@dedene
69NEW	Qwen3.8-Flash-NextIQ3_S	RTX 3080 10GB	52.7	—	Strata	@needmorevram
70NEW	Qwen3.8-Flash-NextIQ3_S	2× AMD MI50 32GB	52	645	Strata	@liucl
71NEW	Qwen3.8-Flash-Next CoderGSQ-RCO	M5 Max	51	—	llama.cpp	@sebastavar
72NEW	Qwen3.8-Flash-Next4bit	DGX Spark	50.2	—	Vontra MLX	@openaiarka
73NEW	Empero Qwen3.8-35B-A3BQ4_K_M	RTX 3060	50	550	llama.cpp	@Oluwaphilemon1
74NEW	Ternary Bonsai 2 27BUncensoredternary	RTX 3060	50	—	llama.cpp	@sudoingX
75NEW	Qwen3.8-Flash-NextIQ3_S	RTX 5080 16GB	50	—	Strata	@rifaniponk
76NEW	Qwen3.8-Flash-NextIQ3_S	RTX 4060 Ti	50	—	Strata	@xiaobei1943
77NEW	Qwen3.5-9B4bit	M4 Pro	49.3	—	rapid-mlx	@rapidmlx
78▼74	Qwen3-8B4bit	M4 Pro	48.3	—	rapid-mlx	@rapidmlx
79NEW	MiMo-V2.6-Distill-Qwen-9BQ5_K_M	RTX 5060	47	1600	llama.cpp	@stfu0911
80NEW	Qwen3.8-Flash-NextFP8	DGX Spark	45	—	—	@Oluwaphilemon1
81NEW	Qwen3.8-27BNVFP4	DGX Spark	42.3	—	TensorFold 0.6.3	@edurdias
82NEW	Qwen3.8-Flash-Nextunspecified	RTX 5090	42.1	—	crow-nest	@crowqnt
83NEW	Qwen3.8-Flash-Next~2.6 bits	M5 Pro 64GB	41.4	—	Sushi	@aartiles24
84NEW	Qwen3.8-27BUD-Q4_K_XL	RTX 4090	40.7	2659.8	llama.cpp	@Oluwaphilemon1
85NEW	Ornith	GTX 1660 SUPER	40	—	—	@mine_craft_bui
86NEW	Qwen3.8-Flash-NextIQ3_S	2× RTX 3060 12GB	40	—	Strata	@xr_ymchn
87NEW	GLM 5.3NVFP4	DGX Spark	40	—	vLLM	@hypermac6502
88NEW	GLM 5.3EXL3 3.0bpw	DGX Spark	38.4	—	TensorFold	@niklaslenz_ai
89▼84	Qwen3.6-35B-A3BUD-IQ3_XXS	M1 Max 32GB	37	350	—	@moonsteroid
90NEW	GLM 5.3NVFP4	DGX Spark	35	—	TensorFold	@majewskizby
91NEW	Qwen3.8-Flash-NextIQ3_S	RTX 3090 + RTX 2070 SUPER	34.2	—	Strata	@Oxtanedotxyz
92NEW	R RED-SNOW 5.3 FlashUncensored2.49 bpw	DGX Spark	33.0	—	ExLlamaV3	@niklaslenz_ai
93NEW	Qwen3.8-Flash-NextQ2_0	Intel Arc B570 10GB	32.4	—	XeStrata	@llm_frontier_jp
94NEW	GLM 5.3 FlashEXL3 2.05bpw	DGX Spark	32.3	—	TensorFold	@NFT_Chen
95NEW	Qwen3.8-Flash-NextNVFP4	DGX Spark	32	—	TensorFold	@openaiarka
96NEW	Qwen3.8-27BNVFP4	DGX Spark	31.8	—	SGLang 0.5.21	@edurdias
97NEW	GLM 5.3Int4/Int8	DGX Spark	30	840	sparkDash	@majewskizby
98NEW	GLM 5.3 FlashEXL3 2.05bpw	DGX Spark	30	—	TensorFold	@filicroval
99NEW	Qwen3.8-27BNVFP4	DGX Spark	28.7	—	vLLM 0.30.0	@edurdias
100NEW	Qwen3.8-Flash-NextIQ2_XS	RTX 3060	27.7	—	Strata	@dec21ai
101NEW	Qwen3.8-Flash-NextIQ3_S	AMD MI50 32GB	26.8	—	Strata	@yokoyoko001
102NEW	Qwen3.8-27BIQ3_XXS	RTX 5060 Ti	23	—	llama.cpp	@fntAInhead
103NEW	Qwen3.8-27BUD-Q2_K_XL	RTX 5060 Ti	19	—	llama.cpp	@fntAInhead
104NEW	GLM 5.3 FlashEXL3	DGX Spark	18.6	—	EXL3	@vcruz305
105NEW	GLM 5.3 FlashNVFP4	DGX Spark	15	—	vLLM	@sudoingX
106NEW	Nex-N2.5-miniQ4_K_M	RTX 5060 Ti	14	—	llama.cpp	@fntAInhead
107NEW	GLM 5.3 FlashQ2	M4 Max 64GB	11.3	—	unspecified	@hokazuya
108NEW	Qwen3.8-27B TurboFCFusionUncensoredIQ2_M	RTX 5060 Ti	10	—	llama.cpp	@fntAInhead
109NEW	REAP50 GLM 5.3 FlashQ2	Radeon RX 9070 XT	10	200	Strata	@syoyo
110NEW	Qwen3.8-Flash-NextQ2_0	Intel Arc B570 10GB	9.0	—	Strata	@llm_frontier_jp
111NEW	Qwen3.8-Flash-NextQ4_K_M	M5 Max	7.8	—	LM Studio	@miya_mos365

Qwen3.8-Flash-Next

DGX Spark

DGX Spark · NVIDIA · FP4-Local

Pair

#54 / 111 in this filter

#13 / 33 on this chip

#24 / 47 of this model

Decode · tok/s

Prefill

Peak

On this chip

This model elsewhere

Lineage

Officialclaimed

Originating sentence

NVIDIA's Qwen3.8-Flash-Next on Spark 1m context 1 million tokens read from SSD (on the Spark) into KV-cache, TTFT just 8.5 seconds (5.5 directly from the cache) and then 65 tok/s! (YaRN setting factor 4.0 over 262,144)

Flags

YaRN 4.0 · 1M context · SSD KV load

Caveat

First-person DGX Spark. Rank is the quoted 65 tok/s after loading 1M into KV. Quant is NVIDIA FP4-Local. Runtime not named beyond the official stack. YaRN 4.0.

Proof

@jtdavies

Get

Get on HF FP4-Local

Used boxes

1 listed DGX Sparkfrom $7,990 · street $8,050Market