whatmodel.sh · find the best AI models you can run

Best AI models to run locally on a 4GB graphics card

A leaderboard of the best AI models you can run locally on a 4GB graphics card (with 32GB of system RAM). Qwen3.8 27B is #1 right now, followed by Ornith 1.5 35B-A3B and KAT-Coder-V2.5-Dev (35B-A3B). Models are ranked by one score that combines 472 published benchmarks. Each model shows the download that fits and the command to run it.

453 models · scores combine 472 benchmarks · updated 2026-10-11 02:20 UTChow it works · find a leaderboard · benchmark your machine

task[all tasks][agent tasks][coding][chat & writing][math][vision & documents][web search]

Models for 32GB RAM · 4GB graphics card → ~4GB fast · local only · reset

+ more

Detect your exact chip or graphics card:

curl -G whatmodel.sh/ -d max=0 --data-urlencode "mac=$(sysctl -n hw.memsize machdep.cpu.brand_string)"Mac

curl -G whatmodel.sh/ -d max=0 --data-urlencode "ram=$(awk '/MemT/{print $2}' /proc/meminfo)" --data-urlencode "nv=$(nvidia-smi --query-gpu=name,memory.total --format=csv,noheader 2>/dev/null)"Linux

curl.exe -G whatmodel.sh/ -d max=0 --data-urlencode "ram=$((gcim Win32_ComputerSystem).TotalPhysicalMemory)" --data-urlencode "nv=$(nvidia-smi --query-gpu=name,memory.total --format=csv,noheader 2>$null)"Windows

$ terminal

ram[8][16][24][32][48][64][96][128][192][256]GB

gpu[apple silicon][graphics card][none]

vram[1][2][4][8][12][16][24][32][48][64][80][96][128][160]GB

price[local][$0.02][$0.1][$0.5][$2][$10][$50][any]max per M

detecting your OS…

Score vs memory to run

     │                               Spark-X2.5 4B                                  Qwen3.8 27B ● ∘      ∘
  50 ┤ · ●───MiniCPM5 2B─●──────────●───────∘──────∘─────∘────∘────∘───∘───∘──∘─────∘─∘────∘─∘──∘·∘∘ ∘ ∘ ∘∘· ·
     │   ∘               ∘          ∘       ∘      ∘     ∘    ∘    ∘   ∘   ∘  ∘  ∘  ∘ ∘  ∘ ∘         ∘ ∘ ∘
   0 ┤ · ∘ · · · · · · · ∘ · · · · ·∘· · · ··· · · · · · ∘ · · · · ∘ · · · · · · ∘ · · · ∘ ∘ · · ··· · · · · ·
     │   ∘
     └───┬──────────────────────────────────┬──────────────────────────┬──────────────────────────┬───────────
        2GB                                5GB                       10GB                       20GB

#	model	score↓	where↓	speed↓	RAM↓	context↓	get it	sources

#	model	score↓	where↓	RAM↓	context↓	get it	sources
59	Qwen3.8 27B · 4-bit Alibaba	73±1	local (slow)	~19GB	262k	Ollama ›	+3
83	Ornith 1.5 35B-A3B · 4-bit DeepReinforce	71±2†	local (slow)	~24GB	262k	llama.cpp ›
88	KAT-Coder-V2.5-Dev (35B-A3B) · 4-bit Kuaishou (Kwaipilot)	70±2†	local (slow)	~24GB	262k	llama.cpp ›
97	Muse Glimmer 30B · 4-bit Meta	70±1	local (slow)	~19GB	131k	llama.cpp ›
99	Qwen3.6 27B · 4-bit Alibaba	69±1	local (slow)	~20GB	262k	Ollama ›	+4
111	Qwen3.5 27B · 4-bit Alibaba	69±0	local (slow)	~19GB	262k	llama.cpp ›	+8
121	Qwen3.6 35B-A3B · 4-bit Alibaba	68±1	local (slow)	~23GB	262k	llama.cpp ›	+4
122	Ornith 1.0 35B (35B-A3B) · 4-bit DeepReinforce	68±1	local (slow)	~24GB	262k	Ollama ›
133	Gemma 4 31B · 4-bit Google	67±1	local (slow)	~21GB	262k	llama.cpp ›	+20
139	Qwen3.5 35B-A3B · 4-bit Alibaba	67±0	local (slow)	~24GB	262k	llama.cpp ›	+12
153	Spark-X2.5 4B · 4-bit iFLYTEK (SparkLLM)	66±1†	local (slow)	~4.2GB	1M	llama.cpp ›
155	Nemotron Cascade 2 30B-A3B · 4-bit NVIDIA	66±1	local (slow)	~22GB	256k	Ollama ›
156	Laguna XS 2.1 (33B-A3B) · 4-bit Poolside	66±2†	local (slow)	~23GB	262k	Ollama ›
161	MiMo-V2.6-Distill-Qwen-9B · 4-bit Xiaomi	65±2†	local (slow)	~8.6GB		llama.cpp ›
164	EXAONE 4.5 33B · 4-bit LG AI Research	65±2†	local (slow)	~24GB	262k	llama.cpp ›
172	Laguna XS.2 · 4-bit Poolside	65±2	local (slow)	~23GB	262k	llama.cpp ›
181	Gemma 4 26B-A4B · 4-bit QAT Google	64±1	local (slow)	~18GB	262k	Ollama ›	+12
186	MiniCPM5 2B · 4-bit OpenBMB	64±1†	local	~3.1GB	131k	llama.cpp ›
192	Granite 4.2 30B · 4-bit IBM	63±1	local (slow)	~20GB	131k	llama.cpp ›
196	Qwen3.5 9B · 4-bit Alibaba	63±1	local (slow)	~7.4GB	262k	llama.cpp ›	+17
197	Seed-OSS 36B Instruct · 4-bit ByteDance Seed	63±1	local (slow)	~25GB	524k	llama.cpp ›
202	Gemma 4 12B · 4-bit Google	62±1	local (slow)	~9.3GB	262k	llama.cpp ›
204	Nemotron 3.5 Lightning 30B-A3B · 4-bit NVIDIA	62±2	local (slow)	~22GB	262k	Ollama ›
211	Qwen3-VL 32B Thinking · 4-bit Alibaba	62±1	local (slow)	~22GB	256k	llama.cpp ›
214	Apriel 1.6 15B Thinker · 4-bit ServiceNow	62±2†	local (slow)	~11GB		llama.cpp ›
222	Devstral Small 2 24B · 4-bit Mistral	61±3†	local (slow)	~17GB	262k	llama.cpp ›
225	Qwen3 30B-A3B Thinking 2507 · 4-bit Alibaba	61±1	local (slow)	~21GB	262k	llama.cpp ›	+8
226	Qwen3.5 4B · 4-bit Alibaba	61±1	local (slow)	~4.3GB	262k	llama.cpp ›	+10
227	Nemotron-Cascade-14B-Thinking · 4-bit NVIDIA	61±1†	local (slow)	~11GB		llama.cpp ›
228	Mellum2.1 12B-A2.5B Thinking · 4-bit JetBrains	60±2†	local (slow)	~10GB		llama.cpp ›
231	Qwen3 30B-A3B Instruct 2507 · 4-bit Alibaba	60±1	local (slow)	~21GB	262k	llama.cpp ›	+4
232	Qwen3-VL 32B Instruct · 4-bit Alibaba	60±1†	local (slow)	~22GB	256k	llama.cpp ›
233	Falcon-H1R 7B · 4-bit TII	60±1†	local (slow)	~6.8GB	262k	llama.cpp ›
239	Granite 4.2 8B · 4-bit IBM	60±1	local (slow)	~7.1GB	131k	llama.cpp ›
241	Qwen3-VL 30B-A3B Thinking · 4-bit Alibaba	60±1†	local (slow)	~21GB	256k	llama.cpp ›
244	QwQ-32B · 4-bit Alibaba	59±1	local (slow)	~22GB	131k	llama.cpp ›	+7
245	OpenReasoning-Nemotron-32B · 4-bit NVIDIA	59±2†	local (slow)	~22GB		llama.cpp ›
247	Gemma 2 9B IT SimPO · 4-bit Princeton NLP	59±1	local (slow)	~7.6GB		llama.cpp ›
248	Gemma 2 9B IT DPO (Princeton) · 4-bit Princeton NLP	59±1	local (slow)	~7.6GB		llama.cpp ›
250	GLM-4.7-Flash · 4-bit Z.ai	59±1	local (slow)	~21GB	200k	llama.cpp ›
251	EXAONE 4.0 32B · 4-bit LG AI Research	59±1	local (slow)	~22GB	131k	llama.cpp ›
252	Phi-4-reasoning-plus 14B · 4-bit Microsoft	59±1	local (slow)	~11GB		llama.cpp ›
253	Nemotron-Cascade-8B · 4-bit NVIDIA	59±1†	local (slow)	~6.8GB		llama.cpp ›
254	Qwen3 14B · 4-bit Alibaba	59±1	local (slow)	~11GB	33k	llama.cpp ›	+6
257	gpt-oss-20b · 4-bit OpenAI	58±1	local (slow)	~14GB	131k	llama.cpp ›	+15
258	Llama 3.3 Nemotron Super 49B v1 · 2-bit NVIDIA	58±2	local (slow)	~22GB		llama.cpp ›
259	Olmo 3.1 32B Think · 4-bit Ai2	58±1	local (slow)	~22GB	66k	llama.cpp ›
261	Phi-4-reasoning 14B · 4-bit Microsoft	58±1	local (slow)	~11GB	33k	llama.cpp ›
262	Olmo 3 32B Think · 4-bit Ai2	58±1	local (slow)	~22GB	66k	llama.cpp ›
263	Magistral Small 1.2 · 4-bit Mistral	58±1	local (slow)	~17GB	131k	llama.cpp ›
264	Nemotron-Cascade-8B-Thinking · 4-bit NVIDIA	58±2†	local (slow)	~6.8GB		llama.cpp ›
265	Qwen3-VL 30B-A3B Instruct · 4-bit Alibaba	58±1†	local (slow)	~21GB	256k	llama.cpp ›
266	Qwen3-VL 8B Thinking · 4-bit Alibaba	58±1	local (slow)	~6.8GB	256k	llama.cpp ›
267	Qwen3 32B · 4-bit Alibaba	58±0	local (slow)	~22GB	131k	llama.cpp ›	+17
268	Qwen3 Coder 30B-A3B Instruct · 4-bit Alibaba	58±1	local (slow)	~21GB	160k	llama.cpp ›
270	Llama 3.3 Nemotron Super 49B v1.5 · 2-bit NVIDIA	58±1	local (slow)	~18GB	131k	llama.cpp ›
271	Nemotron 3 Nano 30B-A3B · 4-bit NVIDIA	58±1	local (slow)	~22GB	262k	Ollama ›	+3
272	Ministral 3 14B Reasoning 2512 · 4-bit Mistral	58±1	local (slow)	~10GB	262k	llama.cpp ›
275	DeepSeek R1 0528 Qwen3 8B · 4-bit DeepSeek	57±1	local (slow)	~6.8GB	128k	llama.cpp ›
276	Granite 4.2 3B · 4-bit IBM	57±1	local (slow)	~3.8GB	131k	llama.cpp ›
281	Qwen2.5-Omni 7B · 4-bit Alibaba	57±2	local (slow)	~6.4GB		llama.cpp ›
283	QED-Nano · 4-bit LM Provers	57±2	local (slow)	~4.1GB	262k	llama.cpp ›
286	Qwen2.5 Coder 32B Instruct · 4-bit Alibaba	57±1	local (slow)	~22GB	131k	llama.cpp ›	+5
287	Qwen3 4B Thinking 2507 · 4-bit Alibaba	57±1	local (slow)	~4.1GB	262k	llama.cpp ›	+2
292	Sky-T1 32B Preview · 4-bit NovaSky (UC Berkeley)	56±5	local (slow)	~22GB		llama.cpp ›
295	Qwen3-VL 4B Thinking · 4-bit Alibaba	56±1	local (slow)	~4.1GB	256k	llama.cpp ›
296	EXAONE Deep 32B · 4-bit LG AI Research	56±2†	local (slow)	~22GB	33k	Ollama ›
297	DeepSeek R1 Distill Qwen 32B · 4-bit DeepSeek	56±1	local (slow)	~22GB	131k	llama.cpp ›	+5
299	Granite 4.1 30B · 4-bit IBM	56±1†	local (slow)	~19GB		Ollama ›
300	Mistral Small 3.2 24B · 4-bit Mistral	56±1	local (slow)	~17GB	128k	llama.cpp ›	+4
302	Qwen3 8B · 4-bit Alibaba	56±1	local (slow)	~6.8GB	33k	llama.cpp ›	+14
303	DeepSeek R1 Distill Qwen 14B · 4-bit DeepSeek	56±1	local (slow)	~11GB	131k	llama.cpp ›	+2
304	Ministral 3 14B · 4-bit Mistral	56±1	local (slow)	~10GB	262k	llama.cpp ›	+1
305	Nemotron Nano 9B v2 · 4-bit NVIDIA	56±1	local (slow)	~7.6GB	131k	llama.cpp ›
311	Ministral 3 8B Reasoning 2512 · 4-bit Mistral	55±1	local (slow)	~7GB	262k	llama.cpp ›
312	Magistral Small 1.0 (2506) · 4-bit Mistral	55±1	local (slow)	~17GB	128k	llama.cpp ›
313	Qwen3 30B-A3B (Apr 2025, hybrid) · 4-bit Alibaba	55±1	local (slow)	~21GB	33k	llama.cpp ›	+10
314	Molmo 2 8B · 4-bit Ai2	55±2	local (slow)	~6.8GB		llama.cpp ›
317	Qwen3 4B Instruct 2507 · 4-bit Alibaba	55±1	local (slow)	~4.1GB	262k	llama.cpp ›	+4
318	Qwen2.5 Coder 14B Instruct · 4-bit Alibaba	55±1	local (slow)	~11GB	131k	llama.cpp ›
319	Qwen3-VL 8B Instruct · 4-bit Alibaba	55±1	local (slow)	~6.8GB	256k	llama.cpp ›
320	Spark-X2.5 1.7B · 4-bit iFLYTEK (SparkLLM)	55±1†	local	~2.7GB	1M	llama.cpp ›
323	Qwen2.5 32B Instruct · 4-bit Alibaba	55±1	local (slow)	~22GB	131k	llama.cpp ›	+11
325	Bielik 11B v3.0 Instruct · 4-bit SpeakLeash	55±2	local (slow)	~8.5GB		llama.cpp ›
326	Olmo 3.1 32B Instruct · 4-bit Ai2	55±2	local (slow)	~22GB	66k	llama.cpp ›
327	OpenThinker 32B · 4-bit Open Thoughts	55±1†	local (slow)	~22GB	33k	llama.cpp ›
328	Magistral Small 1.1 (2507) · 4-bit Mistral	55±2†	local (slow)	~17GB	40k	llama.cpp ›
329	EXAONE Deep 7.8B · 4-bit LG AI Research	55±2†	local (slow)	~6.5GB	33k	llama.cpp ›
331	OpenThinker2 32B · 4-bit Open Thoughts	54±1	local (slow)	~22GB	33k	llama.cpp ›
332	Falcon-H1 34B Instruct · 4-bit TII	54±3	local (slow)	~23GB	262k	llama.cpp ›
333	Llama 3.3 70B Instruct · 2-bit Meta	54±1	local (slow)	~24GB	128k	llama.cpp ›	+21
335	Nemotron 3 Nano 4B · 4-bit NVIDIA	54±1	local (slow)	~4.4GB	262k	llama.cpp ›
336	Granite 4.1 8B · 4-bit IBM	54±1	local (slow)	~7.1GB	131k	Ollama ›
338	Bielik 11B v2.6 Instruct · 4-bit SpeakLeash	54±3	local (slow)	~8.5GB		llama.cpp ›
340	Mellum2 12B-A2.5B Thinking · 4-bit JetBrains	54±1†	local (slow)	~10GB	131k	llama.cpp ›
342	Mistral Small 3.1 24B (2503) · 4-bit Mistral	54±1	local (slow)	~17GB	128k	llama.cpp ›	+7
349	GLM-4 9B Chat 1M · 4-bit Zhipu AI (THUDM)	53±5	local (slow)	~8.1GB	1M	llama.cpp ›
350	Mistral Small 3 24B (2501) · 4-bit Mistral	53±1	local (slow)	~17GB	32k	llama.cpp ›	+6
355	Llama 3.1 Tulu 3 70B DPO · 2-bit Ai2	53±1	local (slow)	~22GB	131k	llama.cpp ›
357	Qwen2.5-VL 32B Instruct · 4-bit Alibaba	53±2	local (slow)	~22GB	33k	llama.cpp ›	+1
359	s1.1 32B · 4-bit Simple Scaling	53±1	local (slow)	~22GB	33k	llama.cpp ›
361	Olmo 3 7B Think · 4-bit Ai2	53±1	local (slow)	~6.2GB	66k	llama.cpp ›
364	Gemma 3 27B · 4-bit Google	53±1	local (slow)	~19GB	131k	llama.cpp ›	+23
365	Phi-4 14B · 4-bit Microsoft	53±1	local (slow)	~11GB	16k	llama.cpp ›	+11
366	Gemma 4 E4B · 4-bit Google	53±1	local (slow)	~7.1GB	131k	llama.cpp ›	+7
367	Ministral 3 8B · 4-bit Mistral	53±1	local (slow)	~7GB	262k	llama.cpp ›	+2
368	Qwen3-VL 4B Instruct · 4-bit Alibaba	53±1	local (slow)	~4.1GB	256k	llama.cpp ›
370	LFM2.5 2.6B · 4-bit Liquid AI	53±1	local	~3.3GB	131k	llama.cpp ›
371	Bielik 11B v2.1 Instruct · 4-bit SpeakLeash	53±3	local (slow)	~8.5GB		llama.cpp ›
374	Qwen2.5-Math 7B Instruct · 4-bit Alibaba	52±2†	local (slow)	~6.4GB	4k	llama.cpp ›
375	LFM2.5 8B-A1B · 4-bit Liquid AI	52±1	local (slow)	~7GB	128k	llama.cpp ›
379	Bielik 11B v2.2 Instruct · 4-bit SpeakLeash	52±3	local (slow)	~8.5GB		llama.cpp ›
380	Granite 4.0 H Small (32B-A9B) · 4-bit IBM	52±1	local (slow)	~22GB	131k	llama.cpp ›	+1
384	Aya Expanse 32B · 4-bit Cohere	52±2	local (slow)	~22GB	128k	llama.cpp ›	+1
385	Gemma 3 12B · 4-bit Google	52±1	local (slow)	~9.2GB	131k	llama.cpp ›	+10
386	Qwen2.5 Coder 7B Instruct · 4-bit Alibaba	52±1	local (slow)	~6.4GB	131k	llama.cpp ›
387	Qwen2.5 14B Instruct · 4-bit Alibaba	52±1	local (slow)	~11GB	131k	llama.cpp ›	+7
388	Cogito v1 Preview Llama 8B · 4-bit Deep Cogito	51±5	local (slow)	~6.6GB	128k	llama.cpp ›
391	OpenCoder-8B-Instruct · 4-bit INF (infly)	51±1	local (slow)	~6.4GB		llama.cpp ›
394	Ministral 3 3B Reasoning 2512 · 4-bit Mistral	51±1	local (slow)	~3.8GB	262k	llama.cpp ›
395	Qwen3 4B · 4-bit Alibaba	51±1	local (slow)	~4.1GB	33k	llama.cpp ›	+8
397	Llama-3-Instruct-8B-SimPO-ExPO · 4-bit Chujie Zheng (ExPO)	51±1	local (slow)	~6.6GB		llama.cpp ›
398	LFM2 24B-A2B · 4-bit Liquid AI	51±1	local (slow)	~17GB	33k	llama.cpp ›
399	AceReason-Nemotron-1.0-14B · 4-bit NVIDIA	51±2†	local (slow)	~11GB		llama.cpp ›
400	Llama-3-Instruct-8B-SimPO · 4-bit Princeton NLP	51±1	local (slow)	~6.6GB		llama.cpp ›
403	Bielik 11B v2.0 Instruct · 4-bit SpeakLeash	51±3	local (slow)	~8.5GB		llama.cpp ›
404	OpenThinker3 7B · 4-bit Open Thoughts	50±1	local (slow)	~6.4GB	33k	llama.cpp ›
405	CodeQwen1.5 7B Chat · 4-bit Alibaba	50±1	local (slow)	~6.4GB	64k	llama.cpp ›
409	EXAONE Deep 2.4B · 4-bit LG AI Research	50±2†	local	~3.2GB	33k	llama.cpp ›
410	Mellum2 12B-A2.5B Instruct · 4-bit JetBrains	50±2†	local (slow)	~10GB	131k	llama.cpp ›
411	OpenCodeInterpreter-DS-33B · 4-bit M-A-P	50±1	local (slow)	~22GB		llama.cpp ›
413	DeepSeek-Coder 33B Instruct · 4-bit DeepSeek	50±1	local (slow)	~22GB	16k	llama.cpp ›	+1
414	Yi 34B 200K (base) · 4-bit 01.AI	50±5	local (slow)	~23GB	200k	llama.cpp ›
415	Gemma 2 27B IT · 4-bit Google	50±1	local (slow)	~19GB	8k	llama.cpp ›	+10
416	Bielik 11B v2.3 Instruct · 4-bit SpeakLeash	50±2	local (slow)	~8.6GB		llama.cpp ›
417	Qwen2.5-VL 7B Instruct · 4-bit Alibaba	50±1	local (slow)	~6.4GB		llama.cpp ›
419	Codestral 22B v0.1 · 4-bit Mistral AI	50±1	local (slow)	~16GB	32k	llama.cpp ›
421	Llama 3 70B Instruct · 2-bit Meta	50±0	local (slow)	~24GB	8k	llama.cpp ›	+17
424	WizardCoder 33B V1.1 · 4-bit WizardLM	50±1	local (slow)	~22GB	16k	llama.cpp ›
425	Mistral Small Instruct 22B (2409) · 4-bit Mistral AI	49±2	local (slow)	~16GB	32k	llama.cpp ›
427	OLMo 2 32B Instruct (0325) · 4-bit Ai2	49±1	local (slow)	~21GB		llama.cpp ›
428	OpenThinker2 7B · 4-bit Open Thoughts	49±2†	local (slow)	~6.4GB	33k	llama.cpp ›
429	Aya 23 35B · 4-bit Cohere For AI	49±3	local (slow)	~24GB	8k	Ollama ›
430	Granite 4.1 3B · 4-bit IBM	49±1†	local (slow)	~3.7GB		Ollama ›
431	OpenCodeInterpreter-DS-6.7B · 4-bit M-A-P	49±1	local (slow)	~5.8GB		llama.cpp ›
432	Artigenz-Coder-DS-6.7B · 4-bit Artigenz	49±1	local (slow)	~5.8GB		llama.cpp ›
433	LFM2.5 1.2B Thinking · 8-bit Liquid AI	49±2	local	~3.5GB	33k	Ollama ›
434	Qwen3-VL 2B Thinking · 4-bit Alibaba	49±1	local	~3.5GB	256k	Ollama ›
435	speechless-codellama-34b-v2.0 · 4-bit uukuguy	49±1	local (slow)	~23GB		llama.cpp ›
436	Pixtral 12B (2409) · 4-bit Mistral	49±2	local (slow)	~9.6GB		llama.cpp ›
438	MegaBeam-Mistral-7B-512k · 4-bit AWS (aws-prototyping)	49±4	local (slow)	~6.1GB		llama.cpp ›
441	AceReason-Nemotron-1.1-7B · 4-bit NVIDIA	49±2†	local (slow)	~6.4GB		llama.cpp ›
443	Yi-Coder 9B Chat · 4-bit 01.AI	49±1	local (slow)	~7.1GB	128k	llama.cpp ›
444	Qwen2-Math 7B Instruct · 4-bit Alibaba	49±2†	local (slow)	~6.1GB	4k	Ollama ›
445	Magicoder-S-DS 6.7B · 4-bit UIUC (ise-uiuc)	49±1	local (slow)	~5.8GB		llama.cpp ›
446	InternLM2.5 7B Chat · 4-bit Shanghai AI Laboratory (InternLM)	49±2	local (slow)	~6.4GB		llama.cpp ›
448	Qwen2.5 7B Instruct · 4-bit Alibaba	48±1	local (slow)	~6.4GB	131k	llama.cpp ›	+12
449	DeepSeek R1 Distill Qwen 7B · 4-bit DeepSeek	48±1	local (slow)	~6.4GB	131k	llama.cpp ›	+6
450	XwinCoder-34B · 4-bit Xwin-LM	48±1	local (slow)	~23GB		llama.cpp ›
451	StarChat2 15B v0.1 · 4-bit Hugging Face	48±1	local (slow)	~12GB	16k	llama.cpp ›
452	Code Millenials 34B · 4-bit Bud Ecosystem	48±1	local (slow)	~23GB		llama.cpp ›
455	T5 Large · 4-bit Google	48±9	local (slow)	~4.6GB	1k	how to run ›
456	WaveCoder-Ultra-6.7B · 4-bit Microsoft	48±1	local (slow)	~5.8GB		llama.cpp ›
460	Gemma 3n E4B · 4-bit Google	48±1	local (slow)	~6.2GB	32k	llama.cpp ›
462	Granite 4.0 H Tiny (7B-A1B) · 4-bit IBM	48±1	local (slow)	~5.9GB	131k	Ollama ›
463	DeepSeek-Coder 7B Instruct v1.5 · 4-bit DeepSeek	48±1	local (slow)	~5.9GB		llama.cpp ›
465	Qwen2.5 Coder 3B Instruct · 4-bit Alibaba	48±1	local	~3.5GB	33k	Ollama ›
466	DeepSeek-Coder 6.7B Instruct · 4-bit DeepSeek	48±1	local (slow)	~5.8GB	16k	llama.cpp ›
467	MiniCPM3-4B · 4-bit OpenBMB	48±2	local (slow)	~4.1GB	32k	llama.cpp ›
468	Bielik 4.5B v3.0 Instruct · 4-bit SpeakLeash	48±2	local (slow)	~4.5GB		llama.cpp ›
469	Qwen3.5 2B · 4-bit Alibaba	47±1	local	~3.5GB	262k	Ollama ›	+7
470	Gemma 2 9B IT · 4-bit Google	47±1	local (slow)	~7.6GB	8k	llama.cpp ›	+14
472	GLM-4 9B Chat · 4-bit Zhipu AI (Z.ai)	47±2	local (slow)	~7.9GB	128k	Ollama ›	+1
473	Yi-1.5 34B (base) · 4-bit 01.AI	47±3	local (slow)	~23GB	4k	llama.cpp ›
474	Phi-4-mini-reasoning 3.8B · 4-bit Microsoft	47±2	local (slow)	~4.1GB	128k	llama.cpp ›
476	WhiteRabbitNeo-33B-v1 · 4-bit WhiteRabbitNeo	47±1	local (slow)	~22GB		llama.cpp ›
477	DeepSeek R1 Distill Llama 8B · 4-bit DeepSeek	47±1	local (slow)	~6.6GB	131k	llama.cpp ›	+1
478	EXAONE 4.0 1.2B · 4-bit LG AI Research	47±2†	local	~2.8GB	66k	llama.cpp ›
480	speechless-coder-ds-6.7b · 4-bit uukuguy	47±1	local (slow)	~5.8GB		llama.cpp ›
481	Gemma 4 E2B · 4-bit Google	47±1	local (slow)	~5.1GB	131k	llama.cpp ›	+6
482	Magicoder-S-CL-7B · 4-bit UIUC (ise-uiuc)	47±1	local (slow)	~5.8GB		llama.cpp ›
483	Aya Expanse 8B · 4-bit Cohere	47±2	local (slow)	~6.9GB	8k	llama.cpp ›
484	Phind CodeLlama 34B v2 · 4-bit Phind	47±1	local (slow)	~23GB	16k	Ollama ›
486	InternLM2.5 7B Chat 1M · 4-bit Shanghai AI Laboratory (InternLM)	47±4	local (slow)	~6.4GB		llama.cpp ›
487	EuroLLM 9B Instruct · 4-bit UTTER (EuroLLM)	47±3	local (slow)	~7.4GB	4k	llama.cpp ›
488	Granite 4.0 H Micro 3B · 4-bit IBM	47±1	local	~3.5GB	131k	Ollama ›
492	MedGemma 4B IT · 4-bit Google	46±3	local (slow)	~4.1GB		llama.cpp ›
493	OLMo 2 13B Instruct (1124) · 4-bit Allen Institute for AI	46±1	local (slow)	~10GB	4k	llama.cpp ›
494	Falcon-H1 7B Instruct · 4-bit TII	46±3	local (slow)	~6.3GB	262k	llama.cpp ›
496	Cogito v1 Preview Llama 3B · 4-bit Deep Cogito	46±4	local (slow)	~3.8GB	128k	llama.cpp ›
498	Hermes 2 Pro Llama 3 8B · 4-bit Nous Research	46±3	local (slow)	~6.6GB	8k	llama.cpp ›
500	Granite 4.0 Micro 3B · 4-bit IBM	46±1	local (slow)	~3.7GB	131k	llama.cpp ›
501	Command R (08-2024) · 4-bit Cohere	46±1	local (slow)	~22GB	128k	llama.cpp ›	+3
504	T5 Base · 4-bit Google	46±9	local	~2.4GB	1k	how to run ›
505	OpenChat 3.5 0106 · 4-bit OpenChat	46±1	local (slow)	~6.1GB	8k	llama.cpp ›	+1
506	Qwen2-VL 7B Instruct · 4-bit Alibaba	46±1	local (slow)	~7.2GB		llama.cpp ›
507	Qwen2.5-Math 1.5B Instruct · 4-bit Alibaba	46±2†	local	~2.5GB	4k	llama.cpp ›
508	Yi 1.5 34B Chat · 4-bit 01.AI	46±1	local (slow)	~23GB	4k	llama.cpp ›	+5
510	Granite 34B Code Instruct (8K) · 4-bit IBM	46±1	local (slow)	~22GB		Ollama ›
511	WizardCoder Python 34B V1.0 · 4-bit WizardLM	46±1	local (slow)	~23GB	16k	Ollama ›
515	Granite 3.3 8B Instruct · 4-bit IBM	46±1	local (slow)	~6.6GB	128k	llama.cpp ›
518	OpenChat 3.5 0106 Gemma · 4-bit OpenChat	46±3	local (slow)	~6.9GB	8k	llama.cpp ›
519	Olmo 3 7B Instruct · 4-bit Ai2	46±1	local (slow)	~6.2GB	66k	llama.cpp ›
520	InternLM2.5 20B Chat · 4-bit Shanghai AI Laboratory (InternLM)	46±2	local (slow)	~14GB	33k	llama.cpp ›
521	StarCoder2-15B-Instruct-v0.1 · 4-bit BigCode	46±1	local (slow)	~11GB	16k	Ollama ›
523	speechless-starcoder2-15b · 4-bit uukuguy	45±1	local (slow)	~12GB		llama.cpp ›
524	Ministral 3 3B · 4-bit Mistral	45±1	local (slow)	~3.7GB	262k	llama.cpp ›	+2
526	OpenThinker 7B · 4-bit Open Thoughts	45±2†	local (slow)	~6.4GB	33k	llama.cpp ›
527	Phi-3 Medium 128K Instruct · 4-bit Microsoft	45±1	local (slow)	~11GB	128k	llama.cpp ›
528	QwQ-32B-Preview · 4-bit Alibaba	45±2	local (slow)	~22GB	33k	llama.cpp ›	+2
529	DeepSeek-Coder-V2 Lite Instruct (16B-A2.4B) · 4-bit DeepSeek	45±1	local (slow)	~12GB	128k	Ollama ›	+1
533	Phi-3-medium-4k-instruct · 4-bit Microsoft	45±1	local (slow)	~11GB	4k	llama.cpp ›	+1
534	Qwen2-Math 1.5B Instruct · 8-bit Alibaba	45±2†	local	~3.2GB	4k	llama.cpp ›
536	Llama 3.1 8B Instruct · 4-bit Meta	45±0	local (slow)	~6.6GB	128k	llama.cpp ›	+25
538	Mistral NeMo Instruct (2407) · 4-bit Mistral	44±1	local (slow)	~9.4GB	128k	llama.cpp ›	+9
540	Phi-3-mini-4k-instruct (June 2024 update) · 4-bit Microsoft	44±1	local (slow)	~4GB		llama.cpp ›
541	Yi 1.5 9B Chat · 4-bit 01.AI	44±1	local (slow)	~7.1GB	4k	llama.cpp ›	+1
542	Command R7B (12-2024) · 4-bit Cohere	44±1	local (slow)	~6.9GB	128k	llama.cpp ›
543	Qwen2-Audio 7B Instruct · 4-bit Alibaba	44±3	local (slow)	~6.5GB		llama.cpp ›
544	Gemma 3 4B · 4-bit Google	44±1	local (slow)	~4.1GB	131k	llama.cpp ›	+11
545	Eurus-2-7B-PRIME · 4-bit PRIME-RL	44±2	local (slow)	~6.4GB	4k	llama.cpp ›
546	Llama 3.1 Tulu 3 8B · 4-bit Ai2	44±1	local (slow)	~6.6GB	131k	llama.cpp ›
547	Qwen2.5-VL 3B Instruct · 4-bit Alibaba	44±1	local	~3.5GB		llama.cpp ›
548	Qwen1.5 32B Chat · 4-bit Alibaba	44±1	local (slow)	~22GB	32k	llama.cpp ›
549	Granite 20B Code Instruct (8K) · 4-bit IBM	44±1	local (slow)	~14GB		Ollama ›
550	Qwen2 7B Instruct · 4-bit Alibaba	44±1	local (slow)	~6.4GB	131k	llama.cpp ›	+4
551	Falcon3 7B Instruct · 4-bit TII	44±3	local (slow)	~6.3GB	32k	llama.cpp ›
553	Qwen3-VL 2B Instruct · 4-bit Alibaba	44±1	local	~3.5GB	256k	Ollama ›
555	Granite 8B Code Instruct (4K) · 4-bit IBM	44±1	local (slow)	~6.6GB		llama.cpp ›
556	Falcon3 10B Instruct · 4-bit TII	44±3	local (slow)	~8.1GB	32k	llama.cpp ›
558	Qwen3 1.7B · 4-bit Alibaba	44±1	local	~3.3GB	33k	Ollama ›	+10
559	Phi-4-mini 3.8B · 4-bit Microsoft	44±1	local (slow)	~4.1GB	128k	llama.cpp ›	+4
560	Ministral 8B (2410) · 4-bit Mistral	44±1	local (slow)	~6.6GB	128k	llama.cpp ›	+6
561	Granite 3.2 8B Instruct · 4-bit IBM	44±1	local (slow)	~6.6GB	131k	llama.cpp ›
563	Llama 3.1 8B (base) · 4-bit Meta	44±3	local (slow)	~6.6GB		llama.cpp ›
564	LFM2.5 1.2B Instruct · 4-bit Liquid AI	44±2	local	~2.3GB	33k	llama.cpp ›
565	Granite 3.0 8B Instruct · 4-bit IBM	43±1	local (slow)	~6.6GB	4k	llama.cpp ›
566	Phi-3.5-mini-instruct · 4-bit Microsoft	43±1	local (slow)	~4GB	128k	llama.cpp ›	+4
567	Hermes 2 Theta Llama 3 8B · 4-bit Nous Research	43±2	local (slow)	~6.6GB	8k	llama.cpp ›
568	Yi 34B Chat · 4-bit 01.AI	43±1	local (slow)	~23GB	4k	llama.cpp ›	+1
571	Gemma 3n E2B IT · 4-bit Google	43±2	local (slow)	~4.7GB	32k	llama.cpp ›
573	T5 Small · 4-bit Google	43±9	local	~1.8GB	1k	how to run ›
574	MiniCPM-V 2.6 · 4-bit OpenBMB	43±2	local (slow)	~6.4GB		llama.cpp ›
575	Qwen2.5 Coder 1.5B Instruct · 8-bit Alibaba	43±1	local	~3.5GB	33k	llama.cpp ›
576	LFM2 2.6B · 4-bit Liquid AI	43±2	local	~3.1GB	33k	llama.cpp ›
577	SmolLM3 3B · 4-bit Hugging Face	43±1	local (slow)	~4GB	66k	llama.cpp ›
578	Granite 3.1 8B Instruct · 4-bit IBM	43±1	local (slow)	~6.6GB	128k	llama.cpp ›
579	Llama 3.2 11B Vision Instruct · 4-bit Meta	43±2	local (slow)	~8.8GB	128k	Ollama ›
580	Qwen2.5 3B Instruct · 4-bit Alibaba	43±1	local (slow)	~3.7GB	33k	llama.cpp ›	+5
581	DeepSeek-Coder 1.3B Instruct · 8-bit DeepSeek	43±1	local	~3GB	16k	Ollama ›
582	Aya 23 8B · 4-bit Cohere For AI	43±3	local (slow)	~6.9GB	8k	llama.cpp ›
583	Phi-3-small-8k-instruct · 16-bit Microsoft	43±2	local (slow)	~17GB		how to run ›
584	CodeGemma 7B IT (v1.0) · 4-bit Google	43±1	local (slow)	~7.1GB		llama.cpp ›
587	Mathstral 7B v0.1 · 4-bit Mistral AI	42±1	local (slow)	~6.1GB	32k	llama.cpp ›
588	Yi-1.5 9B (base) · 4-bit 01.AI	42±2	local (slow)	~7.1GB	4k	llama.cpp ›
590	Llama 3 8B Instruct · 4-bit Meta	42±0	local (slow)	~6.6GB	8k	llama.cpp ›	+19
592	WizardCoder Python 13B V1.0 · 4-bit WizardLM	42±2	local (slow)	~9.8GB	16k	llama.cpp ›
593	Hermes 3 Llama 3.1 8B · 4-bit Nous Research	42±1	local (slow)	~6.6GB	128k	llama.cpp ›
594	Falcon2 11B (base) · 4-bit TII	42±3	local (slow)	~8.2GB	8k	Ollama ›
595	Granite 34B Code Base (8K) · 4-bit IBM	42±1	local (slow)	~24GB		llama.cpp ›
596	speechless-starcoder2-7b · 4-bit uukuguy	42±1	local (slow)	~6.1GB		llama.cpp ›
597	Command R (03-2024) · 4-bit Cohere	42±1	local (slow)	~24GB	128k	llama.cpp ›	+6
599	OLMo 2 7B Instruct (1124) · 4-bit Allen Institute for AI	42±1	local (slow)	~6.2GB	4k	llama.cpp ›
600	WizardCoder-15B-V1.0 · 4-bit WizardLM	42±1	local (slow)	~12GB		llama.cpp ›
602	Phi-3-vision-128k-instruct · 16-bit Microsoft	41±2	local (slow)	~10GB		how to run ›
603	OpenChat 3.5 · 4-bit OpenChat	41±2	local (slow)	~6.1GB	8k	llama.cpp ›
604	InternVL2-4B · 16-bit Shanghai AI Laboratory (OpenGVLab)	41±2	local (slow)	~10GB		how to run ›
605	Qwen1.5 14B Chat · 4-bit Alibaba	41±1	local (slow)	~11GB	32k	Ollama ›	+3
606	Code Llama 34B Python · 4-bit Meta	41±1	local (slow)	~23GB	16k	Ollama ›
607	speechless-coding-7b-16k-tora · 16-bit uukuguy	41±1	local (slow)	~16GB		how to run ›
608	DeepSeek-Coder 33B Base · 4-bit DeepSeek	41±1	local (slow)	~22GB	16k	llama.cpp ›
609	InternLM2 Chat 20B · 4-bit Shanghai AI Laboratory (InternLM)	41±1	local (slow)	~14GB		llama.cpp ›
614	CodeQwen1.5 7B (base) · 4-bit Alibaba	41±1	local (slow)	~6.4GB	66k	llama.cpp ›
615	Granite 4.0 1B · 8-bit IBM	41±1	local	~3.3GB	128k	llama.cpp ›
618	OpenHermes-2.5-Code-290k-13B · 4-bit ajibawa-2023	41±1	local (slow)	~9.8GB		llama.cpp ›
619	Granite 4.0 H 1B · 8-bit IBM	40±1	local	~3.2GB	131k	Ollama ›
620	Llama3-OpenBioLLM-8B · 4-bit Saama AI Labs	40±2	local (slow)	~6.6GB		llama.cpp ›
621	Nous Hermes 2 Mistral 7B DPO · 4-bit Nous Research	40±2	local (slow)	~6.1GB	33k	llama.cpp ›
624	Code Llama 34B Instruct · 4-bit Meta	40±1	local (slow)	~23GB	16k	llama.cpp ›	+1
625	Code Llama 34B (base) · 4-bit Meta	40±1	local (slow)	~23GB	16k	Ollama ›	+1
626	Starling-LM 7B beta · 4-bit Nexusflow	40±1	local (slow)	~6.1GB	8k	llama.cpp ›	+3
627	LFM2 1.2B · 4-bit Liquid AI	40±2	local	~2.3GB	33k	llama.cpp ›
629	LLaVA-v1.6-34B · 4-bit LLaVA	40±2	local (slow)	~24GB	4k	Ollama ›
630	Bielik 1.5B v3.0 Instruct · 8-bit SpeakLeash	40±3	local	~3.3GB		llama.cpp ›
631	OpenHermes 2.5 Mistral 7B · 4-bit Nous Research	39±1	local (slow)	~6.1GB		llama.cpp ›	+2
632	Granite 3.3 2B Instruct · 4-bit IBM	39±1†	local	~3.1GB	128k	llama.cpp ›
633	Phi-3-mini-128k-instruct (Apr 2024) · 16-bit Microsoft	39±2	local (slow)	~9.5GB		how to run ›	+1
634	Mistral 7B Instruct v0.3 · 4-bit Mistral	39±1	local (slow)	~6.1GB	32k	llama.cpp ›	+4
635	Vicuna 33B v1.3 · 4-bit LMSYS	39±1	local (slow)	~23GB	2k	Ollama ›	+2
636	DeepSeek-Coder 6.7B Base · 4-bit DeepSeek	39±1	local (slow)	~5.8GB	16k	llama.cpp ›
637	Starling-LM 7B alpha · 4-bit UC Berkeley (NEST)	39±1	local (slow)	~6.1GB	8k	llama.cpp ›	+2
638	Granite 8B Code Base (4K) · 4-bit IBM	39±1	local (slow)	~6.6GB		llama.cpp ›
639	StarCoder2 15B · 4-bit BigCode	39±1	local (slow)	~12GB	16k	llama.cpp ›
640	speechless-code-mistral-7b-v1.0 · 4-bit uukuguy	39±1	local (slow)	~17GB		how to run ›
641	Falcon-H1 3B Instruct · 4-bit TII	39±4	local	~3.5GB	131k	llama.cpp ›
642	Phi-2 (2.7B) · 4-bit Microsoft	39±1	local	~3.4GB	2k	llama.cpp ›	+2
643	Granite 3.0 2B Instruct · 4-bit IBM	39±1	local	~3.2GB	4k	llama.cpp ›
645	Code Llama 13B Instruct · 4-bit Meta	39±1	local (slow)	~9.8GB	16k	llama.cpp ›
646	MistralHermes-CodePro-7B-v1 · 4-bit beowolx	39±1	local (slow)	~6.1GB		llama.cpp ›
647	Llama-3-Instruct-8B-SimPO-v0.2 · 4-bit Princeton NLP	39±1	local (slow)	~6.6GB		llama.cpp ›
648	Llama 3.2 3B Instruct · 4-bit Meta	38±1	local (slow)	~3.6GB	131k	llama.cpp ›	+11
649	Qwen-14B-Chat · 4-bit Alibaba	38±1	local (slow)	~11GB		llama.cpp ›
651	CodeGemma 7B (v1.0) · 4-bit Google	38±1	local (slow)	~7.1GB		llama.cpp ›
652	Code Llama 13B Python · 4-bit Meta	38±1	local (slow)	~9.8GB	16k	llama.cpp ›
653	Tulu 2 13B · 4-bit Allen Institute for AI	38±2	local (slow)	~9.8GB		llama.cpp ›
654	Phi-3-mini-4k-instruct (Apr 2024) · 4-bit Microsoft	38±2	local (slow)	~4GB		llama.cpp ›
655	Apertus 8B Instruct (2509) · 4-bit Swiss AI	38±4	local (slow)	~6.9GB	66k	llama.cpp ›
657	Granite 3B Code Instruct (2K) · 4-bit IBM	38±2	local (slow)	~3.7GB		llama.cpp ›
658	DeepSeek-V2 Lite Chat (16B-A2.4B) · 4-bit DeepSeek	38±1	local (slow)	~12GB	32k	Ollama ›
659	BioMistral-7B · 4-bit BioMistral	38±3	local (slow)	~6.1GB		llama.cpp ›
660	Qwen3.5 0.8B · 8-bit Alibaba	38±1	local	~3.1GB	262k	Ollama ›	+2
661	Code Tulu 2 7B · 16-bit Allen Institute for AI	38±2†	local (slow)	~16GB		how to run ›
662	SOLAR 10.7B Instruct v1.0 · 4-bit Upstage	37±1	local (slow)	~8.3GB	4k	llama.cpp ›	+2
663	Phi-3.5-vision-instruct · 16-bit Microsoft	37±2	local (slow)	~10GB		how to run ›
664	Tulu 2 DPO 13B · 4-bit Allen Institute for AI	37±1	local (slow)	~9.8GB		llama.cpp ›
665	Yi-1.5 6B (base) · 4-bit 01.AI	37±3	local (slow)	~5.4GB	4k	llama.cpp ›
666	Qwen2-VL 2B Instruct · 4-bit Alibaba	37±1	local	~3.3GB	33k	llama.cpp ›
667	DeepSeek R1 Distill Qwen 1.5B · 8-bit DeepSeek	37±2	local	~3.5GB	131k	llama.cpp ›	+1
668	Gemma 2 2B IT · 4-bit Google	37±1	local	~3.3GB	8k	llama.cpp ›	+4
669	Qwen2.5 Coder 0.5B Instruct · 4-bit Alibaba	37±1	local	~2GB	33k	llama.cpp ›
670	Granite 20B Code Base (8K) · 4-bit IBM	37±1	local (slow)	~15GB		llama.cpp ›
671	Gemma 1.1 7B IT · 4-bit Google	37±1	local (slow)	~7.1GB	8k	llama.cpp ›	+2
672	Code Llama 7B Python · 4-bit Meta	37±1	local (slow)	~5.8GB	16k	llama.cpp ›
673	Dolphin 2.2.1 Mistral 7B · 4-bit Cognitive Computations	37±2	local (slow)	~6.1GB		llama.cpp ›
674	Code Llama 13B (base) · 4-bit Meta	37±1	local (slow)	~9.8GB	16k	llama.cpp ›	+1
675	Qwen1.5 7B Chat · 4-bit Alibaba	37±1	local (slow)	~6.5GB	32k	llama.cpp ›	+4
676	Yi 1.5 6B Chat · 4-bit 01.AI	37±2	local (slow)	~5.4GB	4k	llama.cpp ›
677	Falcon3 3B Instruct · 4-bit TII	37±3	local (slow)	~3.6GB	32k	llama.cpp ›
678	Zephyr 7B alpha · 4-bit Hugging Face	36±1	local (slow)	~6.1GB		llama.cpp ›
679	Vicuna 13B v1.5 · 4-bit LMSYS	36±2	local (slow)	~9.8GB	4k	llama.cpp ›
680	Granite 3.1 2B Instruct · 4-bit IBM	36±1	local	~3.1GB	128k	llama.cpp ›
681	EXAONE 3.5 2.4B Instruct · 4-bit LG AI Research	36±2	local	~3.2GB	33k	llama.cpp ›
683	Code Llama 7B Instruct · 4-bit Meta	36±1	local (slow)	~5.8GB	16k	llama.cpp ›
684	LFM2 700M · 4-bit Liquid AI	36±2	local	~2GB	33k	llama.cpp ›
685	Granite 3.2 2B Instruct · 4-bit IBM	36±1	local	~3.1GB	131k	llama.cpp ›
686	Mistral 7B Instruct v0.2 · 4-bit Mistral	36±1	local (slow)	~6.1GB	32k	llama.cpp ›	+9
687	WizardLM 13B V1.2 · 4-bit Microsoft	36±1	local (slow)	~9.8GB		llama.cpp ›
688	Tulu 2 7B · 4-bit Allen Institute for AI	36±2	local (slow)	~5.8GB		llama.cpp ›
689	xDAN-L1-Chat-RL-v1 · 4-bit xDAN-AI	36±1	local (slow)	~6.1GB		llama.cpp ›
690	Llama 3 8B (base) · 4-bit Meta	36±1	local (slow)	~6.6GB		llama.cpp ›
691	Mistral 7B v0.1 (base) · 4-bit Mistral AI	36±0	local (slow)	~6.1GB	8k	llama.cpp ›	+4
692	Qwen2.5 1.5B Instruct · 8-bit Alibaba	36±1	local	~3.5GB	33k	llama.cpp ›	+6
694	Gemma 7B (base) · 4-bit Google	35±1	local (slow)	~7.1GB	8k	llama.cpp ›	+3
695	Tulu 2 DPO 7B · 4-bit Allen Institute for AI	35±1	local (slow)	~5.8GB		llama.cpp ›
696	Eurus-2-7B-SFT · 4-bit PRIME-RL	35±3†	local (slow)	~6.4GB		llama.cpp ›
698	StarCoder2 7B · 4-bit BigCode	35±1	local (slow)	~6.1GB	16k	llama.cpp ›
699	Granite 3.0 3B-A800M Instruct · 4-bit IBM	35±1	local (slow)	~3.7GB	4k	llama.cpp ›
700	DeepSeek LLM 7B Chat · 4-bit DeepSeek	35±2†	local (slow)	~5.7GB		Ollama ›
701	StarCoder 15.5B · 4-bit BigCode	35±1	local (slow)	~11GB		Ollama ›
702	Granite 3.1 3B-A800M Instruct · 4-bit IBM	35±2†	local (slow)	~3.6GB	131k	llama.cpp ›
703	Llama 2 13B (base) · 4-bit Meta	34±1	local (slow)	~9.3GB	4k	Ollama ›	+3
704	Code Llama 7B (base) · 4-bit Meta	34±1	local (slow)	~5.8GB	16k	llama.cpp ›	+2
705	Nous-Hermes-13b · 4-bit Nous Research	34±2	local (slow)	~9.8GB	2k	llama.cpp ›
706	ChatGLM3 6B 128K · 4-bit Zhipu AI (THUDM)	34±4	local (slow)	~5.8GB	131k	llama.cpp ›
707	Zephyr 7B beta · 4-bit Hugging Face	34±1	local (slow)	~6.1GB		llama.cpp ›	+3
709	Qwen3 0.6B · 8-bit Alibaba	33±1	local	~3GB	33k	Ollama ›	+4
710	Llama 2 13B Chat · 4-bit Meta	33±1	local (slow)	~9.8GB	4k	llama.cpp ›	+9
712	StarCoder2 3B · 4-bit BigCode	33±1	local	~3.4GB	16k	Ollama ›
713	Granite 3B Code Base (2K) · 4-bit IBM	33±1	local (slow)	~3.7GB		llama.cpp ›
715	StarCoderBase 15.5B · 4-bit BigCode	33±1	local (slow)	~12GB		llama.cpp ›
716	Vicuna 7B v1.5 · 4-bit LMSYS	33±1	local (slow)	~5.8GB	4k	llama.cpp ›	+1
718	Stable Code 3B · 4-bit Stability AI	33±1	local	~3.3GB	16k	Ollama ›
719	OLMo 2 1B Instruct (0425) · 8-bit Allen Institute for AI	33±2	local	~3.2GB	4k	llama.cpp ›
720	DeepSeek-Coder 1.3B Base · 8-bit DeepSeek	32±1	local	~3GB	16k	Ollama ›
721	Orca 2 13B · 4-bit Microsoft	32±2	local (slow)	~9.8GB	4k	llama.cpp ›
722	Gemma 7B IT · 4-bit Google	32±1	local (slow)	~7.1GB	8k	llama.cpp ›	+5
723	Mistral 7B Instruct v0.1 · 4-bit Mistral	32±1	local (slow)	~6.1GB	8k	llama.cpp ›	+2
724	Qwen1.5 4B Chat · 4-bit Alibaba	32±2	local (slow)	~4.1GB	32k	llama.cpp ›
725	Ghost 7B Alpha · 4-bit Ghost X	32±2	local (slow)	~6.1GB		llama.cpp ›
726	CodeGemma 2B (v1.0) · 4-bit Google	32±1	local	~3.2GB		Ollama ›
727	Qwen-7B (base) · 4-bit Alibaba	32±1	local (slow)	~6.6GB	33k	llama.cpp ›
728	Phi-1.5 (1.3B) · 8-bit Microsoft	32±1	local	~3.1GB	2k	llama.cpp ›
729	Falcon-H1 1.5B Instruct · 8-bit TII	31±4	local	~3.3GB	131k	llama.cpp ›
730	Flan-T5 XL · 4-bit Google	31±2	local (slow)	~14GB		how to run ›
731	Qwen2 1.5B Instruct · 8-bit Alibaba	31±1	local	~3.2GB	32k	llama.cpp ›	+1
732	Qwen-7B-Chat · 4-bit Alibaba	31±1	local (slow)	~6.6GB	33k	llama.cpp ›
733	Gemma 3 1B · 4-bit Google	31±1	local	~2.7GB	32k	Ollama ›	+7
734	StarCoderBase 7B · 4-bit BigCode	31±1	local (slow)	~6GB		Ollama ›
735	Llama 2 7B (base) · 4-bit Meta	31±1	local (slow)	~5.5GB	4k	Ollama ›	+5
737	LFM2 350M · 4-bit Liquid AI	30±2	local	~1.7GB	33k	llama.cpp ›
738	SmolLM2 1.7B Instruct · 4-bit Hugging Face	30±1	local	~3.1GB	8k	Ollama ›
739	Llama 3.2 1B Instruct · 8-bit Meta	30±1	local	~2.9GB	131k	llama.cpp ›	+13
740	Falcon 7B (base) · 4-bit TII	30±1	local (slow)	~5.9GB	2k	Ollama ›	+1
741	Llama 2 7B Chat · 4-bit Meta	30±1	local (slow)	~5.8GB	4k	llama.cpp ›	+11
742	Baichuan2-7B-Base · 16-bit Baichuan	30±1	local (slow)	~17GB		how to run ›
745	OLMo 7B Instruct · 16-bit Ai2	29±2	local (slow)	~16GB		how to run ›
746	StableLM-3B-4E1T · 4-bit Stability AI	29±1	local	~3.3GB	4k	llama.cpp ›
747	Gemma 2B (base) · 4-bit Google	29±1	local	~3.4GB	8k	Ollama ›	+3
748	OpenLLaMA 7B v2 · 8-bit OpenLM Research	29±1	local (slow)	~9.1GB		llama.cpp ›
749	Gemma 1.1 2B IT · 4-bit Google	29±1	local	~3.2GB	8k	llama.cpp ›	+1
751	Granite 3.1 1B-A400M Instruct · 8-bit IBM	29±2†	local	~3GB	131k	Ollama ›
753	NexusRaven V2 13B · 4-bit Nexusflow	29±3	local (slow)	~9.3GB	16k	Ollama ›
756	GPT-NeoX-20B · 4-bit EleutherAI	28±1	local (slow)	~15GB	2k	llama.cpp ›
759	StarCoderBase 3B · 4-bit BigCode	28±1	local	~3.4GB		Ollama ›
760	Granite 3.0 1B-A400M Instruct · 8-bit IBM	27±1	local	~3GB	4k	llama.cpp ›
761	RedPajama-INCITE 7B Base · 4-bit Together	27±1	local (slow)	~16GB	2k	how to run ›
762	ChatGLM3-6B · 4-bit Zhipu AI (THUDM)	27±2	local (slow)	~5.8GB		llama.cpp ›
763	Granite 4.0 H 350M · 8-bit IBM	27±1†	local	~1.9GB	33k	Ollama ›
764	Flan-T5 Large · 4-bit Google	27±2	local (slow)	~4.8GB		how to run ›
766	Gemma 2B IT · 4-bit Google	27±1	local	~3.1GB	8k	llama.cpp ›	+3
767	EuroLLM 1.7B Instruct · 4-bit UTTER (EuroLLM)	27±3	local	~2.6GB	4k	llama.cpp ›
768	Qwen2.5 0.5B Instruct · 8-bit Alibaba	27±2	local	~2.6GB	33k	Ollama ›	+1
772	Granite 4.0 350M · 16-bit IBM	26±1	local	~2.9GB	32k	Ollama ›
775	Falcon 7B Instruct · 4-bit TII	25±2	local (slow)	~5.9GB		Ollama ›
776	Qwen2 0.5B Instruct · 8-bit Alibaba	25±2	local	~2.6GB	32k	Ollama ›
777	Pythia 12B · 4-bit EleutherAI	25±1	local (slow)	~9.5GB	2k	llama.cpp ›
778	RedPajama-INCITE Base 3B v1 · 4-bit Together	25±3	local	~3.4GB	2k	llama.cpp ›
779	ChatGLM2-6B · 4-bit Zhipu AI (THUDM)	25±2	local (slow)	~15GB		how to run ›
780	GPT-J 6B · 4-bit EleutherAI	25±1	local (slow)	~14GB	2k	how to run ›
781	OpenLLaMA 3B v2 · 4-bit OpenLM Research	25±1	local (slow)	~4.2GB		llama.cpp ›
782	OPT-6.7B · 4-bit Meta	24±1	local (slow)	~16GB	2k	how to run ›
784	Falcon3 1B Instruct · 8-bit TII	24±4	local	~3.4GB	8k	llama.cpp ›
785	StableLM-Base-Alpha 3B v2 · 4-bit Stability AI	24±1	local (slow)	~13GB	4k	how to run ›
786	Qwen1.5 1.8B Chat · 4-bit Alibaba	24±2	local	~2.8GB	33k	llama.cpp ›
787	Pythia 6.9B · 4-bit EleutherAI	23±1	local (slow)	~6.1GB	2k	llama.cpp ›
788	OpenAssistant SFT-4 Pythia 12B · 4-bit LAION (OpenAssistant)	22±1	local (slow)	~9.5GB		llama.cpp ›
789	Pythia 2.8B · 4-bit EleutherAI	21±1	local	~3.4GB	2k	llama.cpp ›
790	BLOOM 7.1B · 4-bit BigScience	21±1	local (slow)	~7.1GB	2k	llama.cpp ›
791	OLMo 1B · 8-bit Allen Institute for AI	21±2†	local	~2.9GB	2k	llama.cpp ›
792	Phi-1 (1.3B) · 8-bit Microsoft	21±4	local	~3.1GB	2k	llama.cpp ›
793	Qwen1.5 0.5B Chat · 4-bit Alibaba	21±3	local	~1.9GB	33k	llama.cpp ›
794	ChatGLM-6B · 4-bit Zhipu AI (THUDM)	20±2	local (slow)	~16GB		how to run ›
795	Pythia 2.8B (deduped) · 4-bit EleutherAI	20±1	local	~3.4GB	2k	llama.cpp ›
796	Flan-T5 Base · 4-bit Google	20±3†	local	~2.5GB		how to run ›
797	Falcon-RW 1B · 16-bit TII	20±2	local (slow)	~4.3GB	2k	how to run ›
798	OPT-2.7B · 4-bit Meta	19±1	local (slow)	~7.1GB	2k	how to run ›
799	SmolLM2 360M Instruct · 4-bit Hugging Face	19±3	local	~2.3GB	8k	Ollama ›
800	GPT-Neo 2.7B · 4-bit EleutherAI	18±1	local (slow)	~13GB	2k	how to run ›	+1
802	Pythia 1.4B · 8-bit EleutherAI	18±2	local	~3.1GB	2k	llama.cpp ›
803	BLOOM 3B · 4-bit BigScience	18±1	local (slow)	~4GB	2k	llama.cpp ›
804	FastChat-T5 3B · 4-bit LMSYS	17±2	local (slow)	~8.5GB		how to run ›
806	OPT-1.3B · 4-bit Meta	16±1	local (slow)	~4.3GB	2k	how to run ›
807	Gemma 3 270M IT · 8-bit Google	15±4	local	~2.4GB	32k	Ollama ›
808	Pythia 1B · 8-bit EleutherAI	14±1	local	~2.7GB	2k	llama.cpp ›
809	StableLM-Tuned-Alpha 7B · 4-bit Stability AI	14±2	local (slow)	~6.9GB		llama.cpp ›
810	BLOOM 1.7B · 4-bit BigScience	14±1	local	~3.1GB	2k	llama.cpp ›
811	Flan-T5 Small · 4-bit Google	14±3†	local	~1.8GB		how to run ›
812	GPT-Neo 1.3B · 4-bit EleutherAI	14±2	local (slow)	~7.1GB	2k	how to run ›
813	GPT-2 XL (1.5B) · 8-bit OpenAI	11±1	local	~3.3GB	1k	llama.cpp ›	+2
814	BLOOM 1.1B · 8-bit BigScience	11±1	local	~3.2GB	2k	llama.cpp ›
815	Pythia 410M · 4-bit EleutherAI	10±2	local	~2.5GB	2k	how to run ›
818	Zyte-1B · 8-bit venkycs (aihub-app)	8±2	local	~2.8GB		llama.cpp ›
819	GPT-Neo 125M · 4-bit EleutherAI	5±5	local	~2.1GB	2k	how to run ›
820	BLOOM 560M · 4-bit BigScience	5±2	local	~2.7GB	2k	how to run ›
821	OPT-350M · 4-bit Meta	5±2	local	~2.2GB	2k	how to run ›
822	StableLM-Base-Alpha 3B · 4-bit Stability AI	4±1	local (slow)	~17GB	4k	how to run ›
823	Pythia 160M · 4-bit EleutherAI	3±2	local	~1.9GB	2k	how to run ›
824	OPT-125M · 4-bit Meta	3±2	local	~1.8GB	2k	how to run ›
825	GPT-2 Large (774M) · 8-bit OpenAI	1±2	local	~2.4GB	1k	llama.cpp ›
826	GPT-2 Medium (355M) · 4-bit OpenAI	-1±2	local	~3.1GB	1k	how to run ›
827	Pythia 70M · 4-bit EleutherAI	-7±2	local	~1.7GB	2k	how to run ›
828	GPT-2 Small (124M) · 4-bit OpenAI	-9±3	local	~2.1GB	1k	how to run ›
—	K2 Horizon 7B · gguf-q4_k_m Institute of Foundation Models (MBZUAI)	not scored yet	local (slow)	~7.7GB	524k	llama.cpp ›

+ 377 more don't fit your machine or budget · show them

+ 163 models with fewer than 3 results left off, since one or two tests say little about overall ability · show them

830 of 1134 catalogue models listed here · sources and licences

* new or thinly tested: ranked on too few benchmark results · † only the maker’s own numbers so far · ~RAM is estimated; full context may need more memory