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

Best AI models to run locally on a 16GB PC without a graphics card

A leaderboard of the best AI models you can run locally on a computer with 16GB of RAM and no graphics card. Spark-X2.5 4B is #1 right now, followed by MiMo-V2.6-Distill-Qwen-9B and MiniCPM5 2B. Models are ranked by one score that combines 575 published benchmarks. Each model shows the download that fits and the command to run it.

Filter by your hardware and budget → ~10GB for models (CPU, slow) · local only · reset

ram[2][4][8][16][24][32][48][64][96][128][192][256]GB

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

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

Filter by task

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

New #1s for your hardware:

orRSS

detecting your OS…

$ terminal

Score vs release date

     │                                                     Gemma 2 9B IT SimPO           Qwen3.5 9B ●────●∘
  50 ┤ · · · · · · · · · · · · · · · · · ·DeepSeek-Coder 6.7B Instr… ●∘●∘●∘●●●●∘∘─∘∘∘∘∘∘∘∘∘∘∘∘∘∘∘─∘∘∘∘∘·∘∘ · ·
     │                                             ●─●──●───∘─∘∘●─●●∘∘∘∘∘∘∘∘∘∘ ∘∘ ∘∘∘∘∘∘∘∘ ∘  ∘     ∘
   0 ┤ · · · · ●─●─∘────────────────∘──────────────∘ ∘ ·∘· ·∘· · · · · ·∘· · · · · · · · · ·∘· · ··· · · · · ·
     │   ●──●──┘                                            ∘
     └───────────────┬────────────────────────┬─────────────────────────┬─────────────────────────┬───────────
                  2020-01                  2022-01                   2024-01                   2026-01

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

#	model	score↓	where↓	RAM↓	speed↓	context↓	get it	sources
158	Spark-X2.5 4B · 4-bit iFLYTEK (SparkLLM)	65±1†	local (slow)	~3.8GB	~192 tok/s	1M	llama.cpp ›
171	MiMo-V2.6-Distill-Qwen-9B · 4-bit Xiaomi	64±2†	local (slow)	~8.6GB			llama.cpp ›
196	MiniCPM5 2B · 4-bit OpenBMB	63±1†	local (slow)	~2.5GB		131k	llama.cpp ›
208	Qwen3.5 9B · 4-bit Alibaba	62±0	local (slow)	~6.7GB	~15 tok/s	262k	llama.cpp ›	+19
241	Qwen3.5 4B · 4-bit Alibaba	60±1	local (slow)	~3.8GB		262k	llama.cpp ›	+11
244	Gemma 2 9B IT SimPO · 4-bit Princeton NLP	60±1	local (slow)	~7.6GB			llama.cpp ›
245	Gemma 2 9B IT DPO (Princeton) · 4-bit Princeton NLP	60±1	local (slow)	~7.6GB			llama.cpp ›
246	Falcon-H1R 7B · 4-bit TII	60±1†	local (slow)	~6GB		262k	llama.cpp ›
253	Granite 4.2 8B · 4-bit IBM	59±1	local (slow)	~7.2GB		131k	llama.cpp ›
270	Nemotron-Cascade-8B · 4-bit NVIDIA	58±1†	local (slow)	~6.8GB			llama.cpp ›
279	Qwen3-VL 8B Thinking · 4-bit Alibaba	58±1	local (slow)	~6.9GB		256k	llama.cpp ›
288	DeepSeek R1 0528 Qwen3 8B · 4-bit DeepSeek	57±1	local (slow)	~6.9GB	~3.6 tok/s	128k	llama.cpp ›
290	Granite 4.2 3B · 4-bit IBM	57±1	local (slow)	~3.4GB		131k	llama.cpp ›
297	Nemotron-Cascade-8B-Thinking · 4-bit NVIDIA	57±2	local (slow)	~6.8GB			llama.cpp ›
299	Qwen2.5-Omni 7B · 4-bit Alibaba	57±2	local (slow)	~5.8GB			llama.cpp ›
305	QED-Nano · 4-bit LM Provers	56±2	local (slow)	~4.1GB		262k	llama.cpp ›
306	Qwen3 4B Thinking 2507 · 4-bit Alibaba	56±1	local (slow)	~4.4GB	~36 tok/s	262k	llama.cpp ›	+3
310	Qwen3-VL 4B Thinking · 4-bit Alibaba	56±1	local (slow)	~4.4GB		256k	llama.cpp ›
318	Qwen3 8B · 4-bit Alibaba	56±1	local (slow)	~6.9GB	~20 tok/s	33k	llama.cpp ›	+15
321	Molmo 2 8B · 4-bit Ai2	56±2	local (slow)	~6.8GB			llama.cpp ›
323	Nemotron Nano 9B v2 · 4-bit NVIDIA	55±1	local (slow)	~7.6GB		131k	llama.cpp ›
325	Qwen3 4B Instruct 2507 · 4-bit Alibaba	55±1	local (slow)	~4.4GB	~60 tok/s	262k	llama.cpp ›	+4
327	Bielik 11B v3.0 Instruct · 4-bit SpeakLeash	55±2	local (slow)	~8.5GB			llama.cpp ›
329	Granite 4.1 8B · 4-bit IBM	55±1	local (slow)	~7.2GB	~70 tok/s	131k	Ollama ›
331	Spark-X2.5 1.7B · 4-bit iFLYTEK (SparkLLM)	55±1†	local (slow)	~1.9GB	~314 tok/s	1M	llama.cpp ›
335	Qwen3-VL 8B Instruct · 4-bit Alibaba	55±1	local (slow)	~6.9GB		256k	llama.cpp ›
340	Bielik 11B v2.6 Instruct · 4-bit SpeakLeash	55±3	local (slow)	~8.5GB			llama.cpp ›
355	Ministral 3 8B Reasoning 2512 · 4-bit Mistral	54±1	local (slow)	~6.9GB		262k	llama.cpp ›
358	Nemotron 3 Nano 4B · 4-bit NVIDIA	54±1	local (slow)	~3.5GB	~104 tok/s	262k	llama.cpp ›
367	GLM-4 9B Chat 1M · 4-bit Zhipu AI (THUDM)	54±4	local (slow)	~8.1GB		1M	llama.cpp ›
368	EXAONE Deep 7.8B · 4-bit LG AI Research	54±2†	local (slow)	~6.5GB		33k	llama.cpp ›
379	Bielik 11B v2.1 Instruct · 4-bit SpeakLeash	53±3	local (slow)	~8.5GB			llama.cpp ›
380	Qwen2.5-Math 7B Instruct · 4-bit Alibaba	53±2†	local (slow)	~5.8GB		4k	llama.cpp ›
381	Olmo 3 7B Think · 4-bit Ai2	53±1	local (slow)	~6.2GB		66k	llama.cpp ›
383	Ministral 3 8B · 4-bit Mistral	53±1	local (slow)	~6.9GB	~68 tok/s	262k	llama.cpp ›	+3
384	Gemma 4 E4B · 4-bit Google	53±1	local (slow)	~6.6GB	~59 tok/s	131k	llama.cpp ›	+8
386	Qwen3-VL 4B Instruct · 4-bit Alibaba	53±1	local (slow)	~4.4GB		256k	llama.cpp ›
390	LFM2.5 2.6B · 4-bit Liquid AI	53±1	local (slow)	~2.4GB		131k	llama.cpp ›
391	Bielik 11B v2.2 Instruct · 4-bit SpeakLeash	53±3	local (slow)	~8.5GB			llama.cpp ›
396	Qwen2.5 Coder 7B Instruct · 4-bit Alibaba	53±1	local (slow)	~5.8GB	~48 tok/s	131k	llama.cpp ›
397	OpenCoder-8B-Instruct · 4-bit INF (infly)	53±1	local (slow)	~6.4GB			llama.cpp ›
398	LFM2.5 8B-A1B · 4-bit Liquid AI	52±1	local (slow)	~5.9GB	~250 tok/s	128k	llama.cpp ›
402	Cogito v1 Preview Llama 8B · 4-bit Deep Cogito	52±4	local (slow)	~6.6GB		128k	llama.cpp ›
403	Llama-3-Instruct-8B-SimPO-ExPO · 4-bit Chujie Zheng (ExPO)	52±1	local (slow)	~6.6GB			llama.cpp ›
404	Llama-3-Instruct-8B-SimPO · 4-bit Princeton NLP	52±1	local (slow)	~6.6GB			llama.cpp ›
409	CodeQwen1.5 7B Chat · 4-bit Alibaba	52±1	local (slow)	~6.4GB		64k	llama.cpp ›
411	Bielik 11B v2.0 Instruct · 4-bit SpeakLeash	51±3	local (slow)	~8.5GB			llama.cpp ›
415	Qwen3 4B · 4-bit Alibaba	51±1	local (slow)	~4.4GB		33k	llama.cpp ›	+8
423	Bielik 11B v2.3 Instruct · 4-bit SpeakLeash	51±2	local (slow)	~8.6GB			llama.cpp ›
427	OpenCodeInterpreter-DS-6.7B · 4-bit M-A-P	51±1	local (slow)	~5.8GB			llama.cpp ›
431	OpenThinker3 7B · 4-bit Open Thoughts	51±1	local (slow)	~6.4GB		33k	llama.cpp ›
432	Granite 4.1 3B · 4-bit IBM	51±1†	local (slow)	~3.7GB	~131 tok/s		Ollama ›
433	Artigenz-Coder-DS-6.7B · 4-bit Artigenz	51±1	local (slow)	~5.8GB			llama.cpp ›
435	Mellum2 12B-A2.5B Instruct · 4-bit JetBrains	51±2†	local (slow)	~9GB		131k	llama.cpp ›
440	Ministral 3 3B Reasoning 2512 · 4-bit Mistral	50±1	local (slow)	~3.6GB		262k	llama.cpp ›
444	Qwen2.5-VL 7B Instruct · 4-bit Alibaba	50±1	local (slow)	~5.8GB			llama.cpp ›
446	Magicoder-S-DS 6.7B · 4-bit UIUC (ise-uiuc)	50±1	local (slow)	~5.8GB			llama.cpp ›
447	EXAONE Deep 2.4B · 4-bit LG AI Research	50±2†	local (slow)	~3.2GB		33k	Ollama ›
448	Pixtral 12B (2409) · 4-bit Mistral	50±2	local (slow)	~9.6GB			llama.cpp ›
451	Qwen2.5 7B Instruct · 4-bit Alibaba	50±1	local (slow)	~5.8GB	~49 tok/s	131k	llama.cpp ›	+12
452	LFM2.5 1.2B Thinking · 4-bit Liquid AI	50±2	local (slow)	~1.3GB		33k	llama.cpp ›
453	Yi-Coder 9B Chat · 4-bit 01.AI	50±1	local (slow)	~6.6GB		128k	llama.cpp ›
454	InternLM2.5 7B Chat · 4-bit Shanghai AI Laboratory (InternLM)	50±2	local (slow)	~6.4GB			llama.cpp ›
457	Qwen2-Math 7B Instruct · 4-bit Alibaba	50±2†	local (slow)	~6.1GB		4k	Ollama ›
458	WaveCoder-Ultra-6.7B · 4-bit Microsoft	50±1	local (slow)	~5.8GB			llama.cpp ›
461	MegaBeam-Mistral-7B-512k · 4-bit AWS (aws-prototyping)	49±4	local (slow)	~5.9GB			llama.cpp ›
464	Qwen3-VL 2B Thinking · 4-bit Alibaba	49±1	local (slow)	~2.7GB		256k	llama.cpp ›
465	Gemma 3n E4B · 4-bit Google	49±1	local (slow)	~5.6GB		32k	llama.cpp ›
467	DeepSeek-Coder 7B Instruct v1.5 · 4-bit DeepSeek	49±1	local (slow)	~5.9GB			llama.cpp ›
468	OpenThinker2 7B · 4-bit Open Thoughts	49±2†	local (slow)	~6.4GB		33k	llama.cpp ›
469	DeepSeek-Coder 6.7B Instruct · 4-bit DeepSeek	49±1	local (slow)	~8.8GB		16k	llama.cpp ›
472	MiniCPM3-4B · 4-bit OpenBMB	49±2	local (slow)	~4.1GB		32k	llama.cpp ›
473	Granite 4.0 H Tiny (7B-A1B) · 4-bit IBM	49±1	local (slow)	~4.8GB	~12 tok/s	131k	Ollama ›
474	Qwen2.5 Coder 3B Instruct · 4-bit Alibaba	49±1	local (slow)	~2.8GB	~9.5 tok/s	33k	Ollama ›
476	Gemma 2 9B IT · 4-bit Google	49±1	local (slow)	~9.5GB		8k	llama.cpp ›	+14
478	Bielik 4.5B v3.0 Instruct · 4-bit SpeakLeash	49±2	local (slow)	~4.5GB			llama.cpp ›
483	GLM-4 9B Chat · 4-bit Zhipu AI (Z.ai)	48±2	local (slow)	~7.1GB		128k	Ollama ›	+1
484	speechless-coder-ds-6.7b · 4-bit uukuguy	48±1	local (slow)	~5.8GB			llama.cpp ›
485	Magicoder-S-CL-7B · 4-bit UIUC (ise-uiuc)	48±1	local (slow)	~5.8GB			llama.cpp ›
487	AceReason-Nemotron-1.1-7B · 4-bit NVIDIA	48±2†	local (slow)	~6.4GB			llama.cpp ›
489	DeepSeek R1 Distill Qwen 7B · 4-bit DeepSeek	48±1	local (slow)	~5.8GB	~64 tok/s	131k	llama.cpp ›	+7
494	Aya Expanse 8B · 4-bit Cohere	48±2	local (slow)	~7GB		8k	llama.cpp ›
495	EuroLLM 9B Instruct · 4-bit UTTER (EuroLLM)	48±2	local (slow)	~7.4GB		4k	llama.cpp ›
498	Granite 4.0 H Micro 3B · 4-bit IBM	48±1	local (slow)	~3.5GB		131k	Ollama ›
503	InternLM2.5 7B Chat 1M · 4-bit Shanghai AI Laboratory (InternLM)	48±4	local (slow)	~6.4GB			llama.cpp ›
504	Falcon-H1 7B Instruct · 4-bit TII	48±3	local (slow)	~6.3GB		262k	llama.cpp ›
508	Qwen3.5 2B · 4-bit Alibaba	47±1	local (slow)	~2.2GB	~174 tok/s	262k	llama.cpp ›	+7
511	Hermes 2 Pro Llama 3 8B · 4-bit Nous Research	47±3	local (slow)	~6.6GB		8k	llama.cpp ›
512	MedGemma 4B IT · 4-bit Google	47±2	local (slow)	~4.1GB			llama.cpp ›
515	OpenChat 3.5 0106 · 4-bit OpenChat	47±1	local (slow)	~6.1GB		8k	llama.cpp ›	+1
519	Granite 4.0 Micro 3B · 4-bit IBM	47±1	local (slow)	~3.7GB	~6.2 tok/s	131k	llama.cpp ›
520	Gemma 4 E2B · 4-bit Google	47±1	local (slow)	~4.5GB		131k	llama.cpp ›	+6
525	Cogito v1 Preview Llama 3B · 4-bit Deep Cogito	47±4	local (slow)	~3.8GB		128k	llama.cpp ›
526	Qwen2.5-Math 1.5B Instruct · 4-bit Alibaba	47±2†	local (slow)	~1.9GB		4k	llama.cpp ›
527	EXAONE 4.0 1.2B · 4-bit LG AI Research	47±1†	local (slow)	~2.8GB		66k	llama.cpp ›
532	Granite 3.3 8B Instruct · 4-bit IBM	47±1	local (slow)	~6.6GB		128k	llama.cpp ›
534	OpenChat 3.5 0106 Gemma · 4-bit OpenChat	47±2	local (slow)	~6.9GB		8k	llama.cpp ›
537	Qwen2-VL 7B Instruct · 4-bit Alibaba	47±1	local (slow)	~6.5GB			llama.cpp ›
541	DeepSeek R1 Distill Llama 8B · 4-bit DeepSeek	46±1	local (slow)	~6.6GB		131k	llama.cpp ›	+2
546	Qwen2-Math 1.5B Instruct · 4-bit Alibaba	46±2†	local (slow)	~2.5GB		4k	Ollama ›
548	Llama 3.1 8B Instruct · 4-bit Meta	46±0	local (slow)	~6.6GB	~35 tok/s	128k	llama.cpp ›	+25
549	Phi-4-mini-reasoning 3.8B · 4-bit Microsoft	46±2	local (slow)	~4.3GB		128k	llama.cpp ›
550	Ministral 3 3B · 4-bit Mistral	46±1	local (slow)	~3.6GB	~128 tok/s	262k	llama.cpp ›	+2
551	Olmo 3 7B Instruct · 4-bit Ai2	46±1	local (slow)	~6.2GB		66k	llama.cpp ›
554	Mistral NeMo Instruct (2407) · 4-bit Mistral	46±1	local (slow)	~9.4GB		128k	llama.cpp ›	+9
556	Llama 3.1 Tulu 3 8B · 4-bit Ai2	46±1	local (slow)	~6.6GB		131k	llama.cpp ›
557	Phi-3-mini-4k-instruct (June 2024 update) · 4-bit Microsoft	46±1	local (slow)	~6GB			llama.cpp ›
558	Yi 1.5 9B Chat · 4-bit 01.AI	46±1	local (slow)	~7.1GB		4k	llama.cpp ›	+1
561	Eurus-2-7B-PRIME · 4-bit PRIME-RL	46±2	local (slow)	~6.4GB		4k	llama.cpp ›
562	Qwen2 7B Instruct · 4-bit Alibaba	46±1	local (slow)	~5.8GB		131k	llama.cpp ›	+4
563	Command R7B (12-2024) · 4-bit Cohere	46±1	local (slow)	~6.9GB		128k	llama.cpp ›
566	Gemma 3 4B · 4-bit Google	45±1	local (slow)	~4.1GB	~74 tok/s	131k	llama.cpp ›	+11
567	Granite 3.0 8B Instruct · 4-bit IBM	45±1	local (slow)	~6.6GB		4k	llama.cpp ›
569	Granite 8B Code Instruct (4K) · 4-bit IBM	45±1	local (slow)	~6.6GB			llama.cpp ›
570	Falcon3 7B Instruct · 4-bit TII	45±2	local (slow)	~6.3GB		32k	llama.cpp ›
571	Phi-4-mini 3.8B · 4-bit Microsoft	45±1	local (slow)	~4.3GB		128k	llama.cpp ›	+4
572	Ministral 8B (2410) · 4-bit Mistral	45±1	local (slow)	~6.6GB		128k	llama.cpp ›	+6
575	Falcon3 10B Instruct · 4-bit TII	45±2	local (slow)	~8.1GB	~13 tok/s	32k	llama.cpp ›
576	Granite 3.2 8B Instruct · 4-bit IBM	45±1	local (slow)	~6.6GB		131k	llama.cpp ›
577	OpenThinker 7B · 4-bit Open Thoughts	45±2†	local (slow)	~6.4GB		33k	llama.cpp ›
579	Hermes 2 Theta Llama 3 8B · 4-bit Nous Research	45±2	local (slow)	~6.6GB		8k	llama.cpp ›
580	Llama 3.1 8B (base) · 4-bit Meta	45±3	local (slow)	~6.6GB			llama.cpp ›
581	Gemma 3n E2B IT · 4-bit Google	45±2	local (slow)	~4.1GB		32k	llama.cpp ›
583	Phi-3.5-mini-instruct · 4-bit Microsoft	45±1	local (slow)	~6GB		128k	llama.cpp ›	+4
584	Qwen2.5-VL 3B Instruct · 4-bit Alibaba	45±1	local (slow)	~2.8GB			llama.cpp ›
585	Qwen3-VL 2B Instruct · 4-bit Alibaba	45±1	local (slow)	~2.7GB		256k	llama.cpp ›
586	LFM2.5 1.2B Instruct · 4-bit Liquid AI	45±2	local (slow)	~1.3GB		33k	llama.cpp ›
587	Qwen3 1.7B · 4-bit Alibaba	45±1	local (slow)	~2.9GB	~96 tok/s	33k	llama.cpp ›	+11
590	Granite 3.1 8B Instruct · 4-bit IBM	44±1	local (slow)	~6.7GB		128k	llama.cpp ›
593	Llama 3.2 11B Vision Instruct · 4-bit Meta	44±2	local (slow)	~8.9GB		128k	Ollama ›
594	Qwen2.5 3B Instruct · 4-bit Alibaba	44±1	local (slow)	~3GB	~61 tok/s	33k	llama.cpp ›	+5
595	LFM2 2.6B · 4-bit Liquid AI	44±1	local (slow)	~2.2GB		33k	llama.cpp ›
599	MiniCPM-V 2.6 · 4-bit OpenBMB	44±2	local (slow)	~5.8GB			llama.cpp ›
600	CodeGemma 7B IT (v1.0) · 4-bit Google	44±1	local (slow)	~7.1GB			llama.cpp ›
601	Yi-1.5 9B (base) · 4-bit 01.AI	44±2	local (slow)	~7.1GB		4k	llama.cpp ›
603	Qwen2-Audio 7B Instruct · 4-bit Alibaba	44±3	local (slow)	~6.5GB			llama.cpp ›
604	DeepSeek-Coder 1.3B Instruct · 4-bit DeepSeek	44±1	local (slow)	~2.9GB		16k	Ollama ›
605	Mathstral 7B v0.1 · 4-bit Mistral AI	44±1	local (slow)	~5.9GB		32k	llama.cpp ›
606	Aya 23 8B · 4-bit Cohere For AI	44±3	local (slow)	~6.9GB		8k	llama.cpp ›
607	Qwen2.5 Coder 1.5B Instruct · 4-bit Alibaba	44±1	local (slow)	~1.9GB		33k	Ollama ›
609	SmolLM3 3B · 4-bit Hugging Face	44±1	local (slow)	~3.5GB		66k	llama.cpp ›
610	Hermes 3 Llama 3.1 8B · 4-bit Nous Research	44±1	local (slow)	~6.6GB		128k	llama.cpp ›
611	Llama 3 8B Instruct · 4-bit Meta	44±0	local (slow)	~6.6GB		8k	llama.cpp ›	+19
615	OpenChat 3.5 · 4-bit OpenChat	43±1	local (slow)	~6.1GB		8k	llama.cpp ›
617	speechless-starcoder2-7b · 4-bit uukuguy	43±1	local (slow)	~6.1GB			llama.cpp ›
618	OLMo 2 7B Instruct (1124) · 4-bit Allen Institute for AI	43±1	local (slow)	~6.2GB		4k	llama.cpp ›
619	Falcon2 11B (base) · 4-bit TII	43±3	local (slow)	~8.2GB		8k	Ollama ›
634	CodeQwen1.5 7B (base) · 4-bit Alibaba	42±1	local (slow)	~6.4GB		66k	llama.cpp ›
635	Nous Hermes 2 Mistral 7B DPO · 4-bit Nous Research	42±3	local (slow)	~6.1GB		33k	llama.cpp ›
638	Granite 4.0 1B · 4-bit IBM	42±1	local (slow)	~2.6GB		128k	llama.cpp ›
640	Granite 4.0 H 1B · 4-bit IBM	42±1	local (slow)	~2.4GB		131k	llama.cpp ›
646	OpenHermes 2.5 Mistral 7B · 4-bit Nous Research	42±1	local (slow)	~5.9GB			llama.cpp ›	+2
649	Starling-LM 7B beta · 4-bit Nexusflow	42±1	local (slow)	~6.1GB		8k	llama.cpp ›	+3
651	Phi-3-mini-128k-instruct (Apr 2024) · 16-bit Microsoft	41±2	local (slow)	~9.5GB			how to run ›	+1
653	Bielik 1.5B v3.0 Instruct · 4-bit SpeakLeash	41±3	local (slow)	~2.6GB			llama.cpp ›
654	Starling-LM 7B alpha · 4-bit UC Berkeley (NEST)	41±1	local (slow)	~6.1GB		8k	llama.cpp ›	+2
655	Granite 3.3 2B Instruct · 4-bit IBM	41±1†	local (slow)	~3.1GB		128k	llama.cpp ›
656	Llama3-OpenBioLLM-8B · 4-bit Saama AI Labs	41±2	local (slow)	~6.6GB			llama.cpp ›
657	Mistral 7B Instruct v0.3 · 4-bit Mistral	41±1	local (slow)	~5.9GB		32k	llama.cpp ›	+4
659	LFM2 1.2B · 4-bit Liquid AI	41±1	local (slow)	~1.3GB		33k	llama.cpp ›
660	DeepSeek-Coder 6.7B Base · 4-bit DeepSeek	41±1	local (slow)	~8.8GB		16k	llama.cpp ›
661	Granite 3.0 2B Instruct · 4-bit IBM	41±1	local (slow)	~3.2GB		4k	llama.cpp ›
663	Llama 3.2 3B Instruct · 4-bit Meta	41±1	local (slow)	~3.5GB	~162 tok/s	131k	llama.cpp ›	+11
664	Granite 8B Code Base (4K) · 4-bit IBM	41±1	local (slow)	~6.6GB			llama.cpp ›
666	Llama-3-Instruct-8B-SimPO-v0.2 · 4-bit Princeton NLP	40±1	local (slow)	~6.6GB			llama.cpp ›
667	Falcon-H1 3B Instruct · 4-bit TII	40±3	local (slow)	~3.5GB		131k	llama.cpp ›
668	Phi-2 (2.7B) · 4-bit Microsoft	40±1	local (slow)	~4.7GB		2k	Ollama ›	+2
674	MistralHermes-CodePro-7B-v1 · 4-bit beowolx	40±1	local (slow)	~6.1GB			llama.cpp ›
676	Phi-3-mini-4k-instruct (Apr 2024) · 4-bit Microsoft	40±2	local (slow)	~6GB			llama.cpp ›
677	CodeGemma 7B (v1.0) · 4-bit Google	40±1	local (slow)	~7.1GB			llama.cpp ›
679	Apertus 8B Instruct (2509) · 4-bit Swiss AI	40±3	local (slow)	~6.9GB		66k	llama.cpp ›
681	Granite 3B Code Instruct (2K) · 4-bit IBM	40±1	local (slow)	~3.7GB			llama.cpp ›
684	BioMistral-7B · 4-bit BioMistral	39±3	local (slow)	~6.1GB			llama.cpp ›
685	SOLAR 10.7B Instruct v1.0 · 4-bit Upstage	39±1	local (slow)	~8.3GB		4k	llama.cpp ›	+2
686	Gemma 2 2B IT · 4-bit Google	39±1	local (slow)	~3.4GB		8k	llama.cpp ›	+4
689	Yi-1.5 6B (base) · 4-bit 01.AI	39±3	local (slow)	~5.4GB		4k	llama.cpp ›
690	Dolphin 2.2.1 Mistral 7B · 4-bit Cognitive Computations	39±2	local (slow)	~5.9GB			llama.cpp ›
691	Gemma 1.1 7B IT · 4-bit Google	39±1	local (slow)	~7.1GB		8k	llama.cpp ›	+2
692	Granite 3.1 2B Instruct · 4-bit IBM	39±1	local (slow)	~3.1GB		128k	llama.cpp ›
695	Code Llama 7B Python · 4-bit Meta	39±1	local (slow)	~5.8GB		16k	llama.cpp ›
696	Granite 3.2 2B Instruct · 4-bit IBM	39±1	local (slow)	~3.1GB		131k	llama.cpp ›
698	Zephyr 7B alpha · 4-bit Hugging Face	38±1	local (slow)	~6.1GB			llama.cpp ›
700	Falcon3 3B Instruct · 4-bit TII	38±3	local (slow)	~3.6GB		32k	llama.cpp ›
701	Qwen2.5 Coder 0.5B Instruct · 4-bit Alibaba	38±1	local (slow)	~1.1GB		33k	Ollama ›
703	Yi 1.5 6B Chat · 4-bit 01.AI	38±2	local (slow)	~4.7GB		4k	llama.cpp ›
705	Qwen2-VL 2B Instruct · 4-bit Alibaba	38±1	local (slow)	~2.6GB		33k	llama.cpp ›
706	Qwen2.5 1.5B Instruct · 4-bit Alibaba	38±1	local (slow)	~2GB	~98 tok/s	33k	llama.cpp ›	+6
707	Tulu 2 7B · 4-bit Allen Institute for AI	38±2	local (slow)	~5.8GB			llama.cpp ›
709	Mistral 7B Instruct v0.2 · 4-bit Mistral	38±1	local (slow)	~5.9GB		32k	llama.cpp ›	+9
710	Code Llama 7B Instruct · 4-bit Meta	38±1	local (slow)	~5.8GB		16k	llama.cpp ›
711	LFM2 700M · 4-bit Liquid AI	38±1	local (slow)	~1GB		33k	llama.cpp ›
712	Qwen3.5 0.8B · 4-bit Alibaba	38±1	local (slow)	~1.4GB	~319 tok/s	262k	llama.cpp ›	+2
714	EXAONE 3.5 2.4B Instruct · 4-bit LG AI Research	38±2	local (slow)	~3.2GB		33k	llama.cpp ›
715	xDAN-L1-Chat-RL-v1 · 4-bit xDAN-AI	37±1	local (slow)	~6.1GB			llama.cpp ›
717	Mistral 7B v0.1 (base) · 4-bit Mistral AI	37±1	local (slow)	~6.1GB		8k	llama.cpp ›	+4
718	Llama 3 8B (base) · 4-bit Meta	37±1	local (slow)	~6.6GB			llama.cpp ›
720	Gemma 7B (base) · 4-bit Google	37±1	local (slow)	~7.1GB		8k	llama.cpp ›	+3
721	Tulu 2 DPO 7B · 4-bit Allen Institute for AI	37±1	local (slow)	~5.8GB			llama.cpp ›
723	Granite 3.1 3B-A800M Instruct · 4-bit IBM	37±2†	local (slow)	~3.6GB	~23 tok/s	131k	llama.cpp ›
724	Granite 3.0 3B-A800M Instruct · 4-bit IBM	37±1	local (slow)	~3.7GB		4k	llama.cpp ›
725	DeepSeek R1 Distill Qwen 1.5B · 4-bit DeepSeek	37±2	local (slow)	~2GB		131k	llama.cpp ›	+1
726	StarCoder2 7B · 4-bit BigCode	37±1	local (slow)	~5.4GB		16k	llama.cpp ›
728	DeepSeek LLM 7B Chat · 4-bit DeepSeek	37±2†	local (slow)	~5.7GB			Ollama ›
732	Eurus-2-7B-SFT · 4-bit PRIME-RL	36±3†	local (slow)	~6.4GB			llama.cpp ›
733	Qwen3 0.6B · 4-bit Alibaba	36±1	local (slow)	~2.1GB	~61 tok/s	33k	llama.cpp ›	+5
735	Code Llama 7B (base) · 4-bit Meta	36±1	local (slow)	~5.8GB		16k	llama.cpp ›	+2
736	Llama 2 13B (base) · 4-bit Meta	36±1	local (slow)	~9.3GB		4k	Ollama ›	+4
738	Zephyr 7B beta · 4-bit Hugging Face	35±1	local (slow)	~6.1GB			llama.cpp ›	+3
740	StarCoder2 3B · 4-bit BigCode	35±1	local (slow)	~3.4GB		16k	llama.cpp ›
741	Granite 3B Code Base (2K) · 4-bit IBM	35±1	local (slow)	~3.7GB			llama.cpp ›
743	Vicuna 7B v1.5 · 4-bit LMSYS	35±1	local (slow)	~5.8GB		4k	llama.cpp ›	+1
744	ChatGLM3 6B 128K · 4-bit Zhipu AI (THUDM)	35±4	local (slow)	~5.8GB		131k	llama.cpp ›
746	OLMo 2 1B Instruct (0425) · 4-bit Allen Institute for AI	35±1	local (slow)	~2.4GB		4k	llama.cpp ›
747	Vicuna 7B v1.3 · 4-bit LMSYS	35±2	local (slow)	~5.8GB		2k	llama.cpp ›
748	Stable Code 3B · 4-bit Stability AI	35±1	local (slow)	~4.8GB		16k	llama.cpp ›
751	DeepSeek-Coder 1.3B Base · 4-bit DeepSeek	34±1	local (slow)	~2.4GB		16k	Ollama ›
752	Qwen1.5 4B Chat · 4-bit Alibaba	34±2	local (slow)	~4.1GB		32k	llama.cpp ›
753	Mistral 7B Instruct v0.1 · 4-bit Mistral	34±1	local (slow)	~5.9GB		8k	llama.cpp ›	+3
754	Gemma 7B IT · 4-bit Google	34±1	local (slow)	~7.1GB		8k	llama.cpp ›	+5
757	CodeGemma 2B (v1.0) · 4-bit Google	34±1	local (slow)	~3.2GB			llama.cpp ›
758	Qwen2 1.5B Instruct · 4-bit Alibaba	34±1	local (slow)	~1.9GB		32k	llama.cpp ›	+1
759	Falcon-H1 1.5B Instruct · 4-bit TII	33±4	local (slow)	~2.4GB		131k	llama.cpp ›
760	Ghost 7B Alpha · 4-bit Ghost X	33±2	local (slow)	~6.1GB			llama.cpp ›
761	Gemma 3 1B · 4-bit Google	33±1	local (slow)	~1.8GB	~180 tok/s	32k	llama.cpp ›	+7
762	SmolLM2 1.7B Instruct · 4-bit Hugging Face	33±1	local (slow)	~3.1GB		8k	llama.cpp ›
763	LFM2 350M · 4-bit Liquid AI	33±1	local (slow)	~1.7GB		33k	llama.cpp ›
764	Qwen-7B (base) · 4-bit Alibaba	33±1	local (slow)	~6.6GB		33k	llama.cpp ›
767	StarCoderBase 7B · 4-bit BigCode	33±1	local (slow)	~6GB			Ollama ›
769	Llama 3.2 1B Instruct · 4-bit Meta	33±1	local (slow)	~1.7GB	~251 tok/s	131k	llama.cpp ›	+13
770	Phi-1.5 (1.3B) · 4-bit Microsoft	33±1	local (slow)	~2.4GB		2k	llama.cpp ›
772	Qwen-7B-Chat · 4-bit Alibaba	32±1	local (slow)	~6.6GB		33k	llama.cpp ›
773	Llama 2 7B Chat · 4-bit Meta	32±1	local (slow)	~8.8GB		4k	llama.cpp ›	+13
776	Llama 2 7B (base) · 4-bit Meta	32±1	local (slow)	~8.5GB		4k	Ollama ›	+6
778	Flan-T5 Large · 4-bit Google	32±1	local (slow)	~4.8GB			how to run ›	+5
779	Granite 3.1 1B-A400M Instruct · 4-bit IBM	31±2†	local (slow)	~2.3GB	~20 tok/s	131k	llama.cpp ›
781	Gemma 2B (base) · 4-bit Google	31±1	local (slow)	~3.2GB		8k	llama.cpp ›	+3
782	Gemma 1.1 2B IT · 4-bit Google	31±1	local (slow)	~3.2GB		8k	llama.cpp ›	+1
783	Falcon 7B (base) · 4-bit TII	31±1	local (slow)	~5.9GB		2k	Ollama ›	+4
787	NexusRaven V2 13B · 4-bit Nexusflow	30±3	local (slow)	~9.3GB		16k	Ollama ›
789	ChatGLM3-6B · 4-bit Zhipu AI (THUDM)	30±2	local (slow)	~5.8GB			llama.cpp ›
790	Qwen2.5 0.5B Instruct · 4-bit Alibaba	30±2	local (slow)	~1.2GB		33k	llama.cpp ›	+1
792	OpenLLaMA 7B v2 · 8-bit OpenLM Research	30±1	local (slow)	~9.1GB			llama.cpp ›
793	StableLM-3B-4E1T · 4-bit Stability AI	30±1	local (slow)	~3.3GB		4k	llama.cpp ›
794	StarCoderBase 3B · 4-bit BigCode	30±1	local (slow)	~3.4GB			Ollama ›
797	Granite 3.0 1B-A400M Instruct · 4-bit IBM	30±1	local (slow)	~2.3GB		4k	llama.cpp ›
801	EuroLLM 1.7B Instruct · 4-bit UTTER (EuroLLM)	29±3	local (slow)	~2.6GB		4k	llama.cpp ›
802	Granite 4.0 H 350M · 4-bit IBM	29±1†	local (slow)	~1.7GB		33k	llama.cpp ›
804	Gemma 2B IT · 4-bit Google	29±1	local (slow)	~3.1GB		8k	llama.cpp ›	+3
806	Granite 4.0 350M · 4-bit IBM	28±1	local (slow)	~1.7GB		32k	llama.cpp ›
810	Qwen2 0.5B Instruct · 4-bit Alibaba	28±2	local (slow)	~1.1GB		32k	llama.cpp ›
816	Falcon3 1B Instruct · 4-bit TII	26±4	local (slow)	~2.7GB		8k	llama.cpp ›
817	Qwen1.5 1.8B Chat · 4-bit Alibaba	26±2	local (slow)	~2.7GB		33k	Ollama ›
820	Falcon 7B Instruct · 4-bit TII	26±2	local (slow)	~5.9GB			Ollama ›	+1
822	OpenAssistant SFT-4 Pythia 12B · 4-bit LAION (OpenAssistant)	25±1	local (slow)	~9.5GB			llama.cpp ›	+1
823	Flan-T5 Base · 4-bit Google	25±1	local (slow)	~2.5GB			how to run ›	+3
824	RedPajama-INCITE Base 3B v1 · 4-bit Together	25±3	local (slow)	~3.4GB		2k	llama.cpp ›
826	Pythia 12B · 4-bit EleutherAI	25±1	local (slow)	~9.5GB		2k	llama.cpp ›	+1
827	OpenLLaMA 3B v2 · 4-bit OpenLM Research	25±1	local (slow)	~4.2GB			llama.cpp ›
834	Qwen1.5 0.5B Chat · 4-bit Alibaba	23±3	local (slow)	~1.8GB		33k	Ollama ›
835	Pythia 6.9B · 4-bit EleutherAI	23±1	local (slow)	~6.1GB		2k	llama.cpp ›	+2
837	Phi-1 (1.3B) · 4-bit Microsoft	22±4	local (slow)	~2.4GB		2k	llama.cpp ›
839	SmolLM2 360M Instruct · 8-bit Hugging Face	22±3	local (slow)	~1.2GB		8k	llama.cpp ›
841	OLMo 1B · 4-bit Allen Institute for AI	21±2†	local (slow)	~2.2GB		2k	llama.cpp ›
842	BLOOM 7.1B · 4-bit BigScience	21±1	local (slow)	~7.1GB		2k	llama.cpp ›	+7
843	Pythia 2.8B · 4-bit EleutherAI	20±1	local (slow)	~3.4GB		2k	llama.cpp ›
844	Pythia 2.8B (deduped) · 4-bit EleutherAI	20±1	local (slow)	~3.4GB		2k	llama.cpp ›
845	Falcon-RW 1B · 16-bit TII	20±2	local (slow)	~4.3GB		2k	how to run ›
846	OPT-2.7B · 4-bit Meta	19±1	local (slow)	~7.1GB		2k	how to run ›	+3
847	StableLM-Tuned-Alpha 7B · 4-bit Stability AI	18±2	local (slow)	~6.9GB			llama.cpp ›
850	T5 Large · 4-bit Google	17±2	local (slow)	~4.6GB		1k	how to run ›
851	Gemma 3 270M IT · 8-bit Google	17±4	local (slow)	~1.2GB		32k	llama.cpp ›
852	Flan-T5 Small · 4-bit Google	16±3	local (slow)	~1.8GB			how to run ›
853	BLOOM 3B · 4-bit BigScience	16±1	local (slow)	~4GB		2k	llama.cpp ›	+2
856	OPT-1.3B · 4-bit Meta	14±1	local (slow)	~4.3GB		2k	how to run ›	+5
857	T5 Base · 4-bit Google	14±3	local (slow)	~2.4GB		1k	how to run ›
858	Pythia 1.4B · 4-bit EleutherAI	14±1	local (slow)	~2.4GB		2k	llama.cpp ›
859	Pythia 1B · 4-bit EleutherAI	12±1	local (slow)	~2.2GB		2k	llama.cpp ›
862	GPT-Neo 1.3B · 4-bit EleutherAI	12±2	local (slow)	~7.1GB		2k	how to run ›	+3
863	T5 Small · 4-bit Google	12±3	local (slow)	~1.8GB		1k	how to run ›
864	BLOOM 1.7B · 4-bit BigScience	11±1	local (slow)	~3.1GB		2k	llama.cpp ›	+1
865	Zyte-1B · 4-bit venkycs (aihub-app)	11±2	local (slow)	~2.2GB			llama.cpp ›
868	Pythia 410M · 4-bit EleutherAI	8±2	local (slow)	~1.8GB		2k	llama.cpp ›
870	BLOOM 1.1B · 4-bit BigScience	7±1	local (slow)	~2.6GB		2k	llama.cpp ›
871	GPT-2 XL (1.5B) · 4-bit OpenAI	7±1	local (slow)	~2.8GB	~75 tok/s	1k	llama.cpp ›	+9
874	OPT-350M · 4-bit Meta	1±1	local (slow)	~2.2GB		2k	how to run ›	+1
877	BLOOM 560M · 4-bit BigScience	0±1	local (slow)	~2.1GB		2k	llama.cpp ›	+1
878	Pythia 160M · 4-bit EleutherAI	-2±2	local (slow)	~1.6GB		2k	llama.cpp ›
879	GPT-2 Large (774M) · 4-bit OpenAI	-3±1	local (slow)	~2GB	~137 tok/s	1k	llama.cpp ›	+3
880	OPT-125M · 4-bit Meta	-3±1	local (slow)	~1.8GB		2k	how to run ›	+1
881	GPT-Neo 125M · 4-bit EleutherAI	-3±2	local (slow)	~2.1GB		2k	how to run ›
882	GPT-2 Medium (355M) · 4-bit OpenAI	-5±1	local (slow)	~1.8GB		1k	llama.cpp ›	+3
883	Pythia 70M · 4-bit EleutherAI	-13±2	local (slow)	~1.6GB		2k	llama.cpp ›
884	GPT-2 Small (124M) · 4-bit OpenAI	-13±2	local (slow)	~1.6GB		1k	llama.cpp ›	+3
—	K2 Horizon 7B · gguf-q4_k_m Institute of Foundation Models (MBZUAI)	not scored yet	local (slow)	~8GB		524k	llama.cpp ›
—	LFM3 1.7B · q4_k_m Liquid AI	not yet calibrated	local (slow)	~2.6GB		262k	llama.cpp ›

+ 624 more don't fit your machine or budget · show them

+ 111 models with fewer than 3 results left off, since one or two tests say little about overall ability · show them

919 of 1227 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