Models

Everytrackedmodel.

One page per model: what it needs in memory, where it runs, and how it moves on the radar — all from measured readings and the same fit engine as the finder.

● Collected 24 min agoThe ranked board →The radar view →

Every tracked model — memory, fit and momentum
#	Model	Measured file	Heat	Runs comfortably at 8K
1	Qwen3.8 Flash-NextAlibaba Qwen · general, coding, agents	180B MoE · Q4_K · 103.7 GB	85	0 of 14
2	Llama 3.1 8BMeta · general, coding	8B · Q4_K_M · 4.6 GB	85	unknown · attention shape unpublished
3	Gemma 4 31BGoogle · general, research	31.3B · Q4_K_M · 17.1 GB	80	5 of 14
4	Qwen3.8 27BAlibaba Qwen · general, coding, agents, research	27.8B · Q4_K_M · 15.3 GB	78	7 of 14
5	Llama 3.2 3BMeta · general	3B · Q4_K_M · 1.9 GB	78	unknown · attention shape unpublished
6	Qwen3 8BAlibaba Qwen · general, coding, agents	8.2B · Q4_K_M · 4.7 GB	77	14 of 14
7	GLM-5.3-FlashZhipu AI · general, coding, agents	321B · Q4_K · 186 GB	76	0 of 14
8	Gemma 4 26B-A4BGoogle · general, research	25.8B MoE · Q4_K_M · 15.8 GB	74	7 of 14
9	Ornith 1.5 35B-A3BOrnith AI · general, coding, agents	36B MoE · Q4_K_M · 20.2 GB	71	5 of 14
10	gpt-oss 20BOpenAI · general, agents, research	20.9B MoE · Q4_K_M · 10.8 GB	71	12 of 14
11	GLM-5.3Zhipu AI · general, agents, research	753B · Q4_K · 435.2 GB	70	0 of 14
12	Qwen3 0.6BAlibaba Qwen · general	0.6B · Q8_0 · 0.6 GB	69	14 of 14
13	Ornith 1.5 9BOrnith AI · general, coding	9.7B · Q4_K_M · 5.4 GB	68	14 of 14
14	GLM-5.2Zhipu AI · coding, agents, research	753B MoE · Q4_K_M · 433.8 GB	67	0 of 14
15	gpt-oss 120BOpenAI · general, research, agents	117B MoE · Q4_K_M · 58.5 GB	65	1 of 14
16	Qwen3-Coder NextAlibaba Qwen · coding, agents	79.7B MoE · Q4_K_M · 45.1 GB	62	2 of 14
17	Llama 3.3 70BMeta · general, research, agents	70.6B · Q4_K_M · 39.6 GB	58	2 of 14
18	Gemma 4 E4BGoogle · general	8B · Q4_K_M · 4.6 GB	57	14 of 14
19	DeepSeek-V4-FlashDeepSeek · general, coding, research	304B MoE · Q4_K · 144.4 GB	55	0 of 14
20	Qwen3 4BAlibaba Qwen · general, coding	4B · Q4_K_M · 2.3 GB	52	14 of 14
21	Gemma 4 12BGoogle · general, research	12B · Q4_K_M · 6.6 GB	51	14 of 14
22	Qwen3.8 2.4T-A95BAlibaba Qwen · general, agents, research	2.4T-A95B · IQ4_XS · 1220.8 GB	50	0 of 14
23	Granite 4.2 3BIBM · general, agents	3.7B · Q4_K_M · 2.1 GB	48	14 of 14
24	Gemma 3 27BGoogle · general, research	27.4B · Q4_K_M · 15.4 GB	46	7 of 14
25	Granite 4.2 8BIBM · general, coding, agents	8.8B · Q4_K_M · 5 GB	46	14 of 14
26	GLM-4.7-FlashZhipu AI · coding, agents	31.2B MoE · Q4_K_M · 17.1 GB	40	7 of 14
27	Ornith 1.5 397BOrnith AI · general, coding, research	403B MoE · Q4_K_M · 227.5 GB	40	0 of 14
28	Qwen3 14BAlibaba Qwen · general, coding, research	14.8B · Q4_K_M · 8.4 GB	39	12 of 14
29	Granite 4.2 30BIBM · general, coding, agents	29.3B · Q4_K_M · 16.5 GB	39	5 of 14
30	Mistral Small 3.2Mistral AI · general, agents	24B · Q4_K_M · 13.3 GB	37	7 of 14
31	DeepSeek-V4-ProDeepSeek · research, coding, agents	1650B MoE · Q4_K · 791.3 GB	35	0 of 14
32	Hy3Tencent · general, research	299B MoE · Q4_K_M · 166.3 GB	34	0 of 14
33	Qwen3-Coder 30B-A3BAlibaba Qwen · coding, agents	30.5B MoE · Q4_K_M · 17.3 GB	33	5 of 14
34	Devstral Small 2 24BMistral AI · coding, agents	24B · Q4_K_M · 13.3 GB	32	7 of 14
35	KAT-Coder V2.5Kwaipilot · coding, agents	34.7B MoE · Q4_K_M · 19.9 GB	30	5 of 14
36	LFM2.5 8B-A1BLiquid AI · general, agents	8.5B MoE · Q4_K_M · 4.8 GB	28	14 of 14
37	Phi-4 MiniMicrosoft · general	3.8B · Q4_K_M · 2.3 GB	28	14 of 14
38	GLM-4.5-AirZhipu AI · coding, agents, general, research	110B MoE · Q4_K_M · 68 GB	28	1 of 14
39	LFM2.5 2.6BLiquid AI · general, agents	2.7B · Q4_K_M · 1.6 GB	27	14 of 14
40	Ling 3.0 TinyInclusionAI · general, agents	7.9B MoE · Q4_K_M · 4.5 GB	26	14 of 14
41	Qwen3 32BAlibaba Qwen · general, coding, research, agents	32.8B · Q4_K_M · 18.4 GB	26	5 of 14
42	Ling 3.0 FlashInclusionAI · general, agents	127B MoE · Q4_K_M · 70.1 GB	25	1 of 14
43	Qwen AgentWorld 35B-A3BAlibaba Qwen · agents, coding	34.7B MoE · Q4_K_M · 20.6 GB	25	5 of 14
44	LFM2.5 1.2BLiquid AI · general	1.2B · Q4_K_M · 0.7 GB	18	14 of 14
45	Nemotron Nano 9B v2NVIDIA · general, research	8.9B · Q4_K_M · 6.1 GB	18	14 of 14
46	DeepSeek-R1DeepSeek · research	685B MoE · Q4_K_M · 376.7 GB	collecting	0 of 14

Comfortable = EXCELLENT or GOOD on that device at 8K context. Models without a measured open-weights file show no fit.