Models

Name: OpenWeights Terminal measured pairs
Creator: Dvidia
License: https://owterminal.com/cite

Speed is not a model property. Rank is the fastest measured pair of that family.

#	Model · Build	Hardware	Decode · tok/s	Get
1	Qwen3.5 2B	RTX 5090 32GB	351
2	Swift 1.5 Qwen3.8 Flash-Next	RTX 4090 24GB	220
3	Qwen3.8-27B	RTX A5000	177.6	HF
4	Gemma-4 26B-A4B-it	RTX 5090 32GB	173
5	Qwen3.8-Flash-Next	RTX 5090 32GB	142
6	Qwen3.6-35B-A3B abliterated	2× DGX Spark	94.4
7	Qwen3.6-35B-A3B	M4 Pro	85.5	HF
8	Qwen3.5-4B	M4 Pro	82.8
9	AliceAI-Foundation-80B-A3B	M5 Max 128GB	65	HF
10	GLM 5.3 Flash	2× DGX Spark	60.4
11	Empero Qwen3.8-35B-A3B	RTX 3060 12GB	50
12	Qwen3.5-9B	M4 Pro	49.3
13	Qwen3-8B	M4 Pro	48.3
14	MiMo-V2.6-Distill-Qwen-9B	RTX 5060 8GB	47
15	Ornith	GTX 1660 SUPER 6GB	40	HF
16	GLM 5.3	4x DGX Spark	30
17	Nex-N2.5-mini	RTX 5060 Ti 16GB	14
18	Qwen3.8-27B TurboFC	RTX 5060 Ti 16GB	10
19	Qwen3.8-27B Mythos	—	—	HF
20	Qwen3.8-35B-A3B APEX	—	—	HF
21	Huihui Qwen3.8-27B	—	—	HF
22	Qwen3.8-27B OBLITERATED	—	—	HF
23	Qwen3.8-27B Uncensored	—	—	HF
24	Qwen3.8-27B Heretic	—	—	HF
25	Qwen3.8-27B Splash	—	—	HF
26	SuperQwen3.8-27B	—	—	HF
27	Flash-Next RVN	—	—	HF
28	Qwen3.8-27B Heretic ARA	—	—	HF
29	Qwen3.8-27B Coletti	—	—	HF
30	Qwen3.8-27B NoRefusal	—	—	HF
31	Qwen3.8-27B KCRN	—	—	HF
32	Qwen3.8-27B Fable	—	—	HF
33	Qwen3.8-9B Heretic	—	—	HF
34	DeepSeek v4.1 Flash abliterated	—	—	HF
35	Gemma-4 26B Abliterix	—	—	HF
36	Gemma-4 26B Uncensored	—	—	HF
37	Gemma-4 31B Uncensored	—	—	HF
38	Gemma-4 12B OBLITERATED	—	—	HF
39	Qwen3.8-27B Ultra Heretic	—	—	HF
40	Qwen3.8-27B Aggressive	—	—	HF
41	Qwen3.8-27B Twin Turbo	—	—	HF
42	GLM 5.3 Flash abliterated	—	—	HF
43	GLM 5.3 Flash Uncensored	—	—	HF
44	GLM 5.3 Uncensored	—	—	HF
45	GLM 5.3 EXL3 abliterated	—	—	HF
46	DeepSeek v4.1 Flash	—	—
47	MiMo-V2.6-Flash	—	—
48	Ling-3.0-Flash	—	—	HF
49	Nemotron 3.5 Lightning	—	—
50	GLM-5.2	—	—
51	Kimi K3	—	—	HF
52	CYBER-FROST 3.8	—	—	HF
53	Qwen-Image-2.1	—	—	HF
54	Qwen-Image-2.1 Uncensored	—	—	HF
55	Ternary Bonsai 2 27B	—	—	HF
56	MiniMax-M2.5	—	—