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

Best AI models for chat and writing to run locally

A leaderboard of the best AI models for chat and writing you can run on your own computer. Kimi K3 is #1 right now, followed by Qwen3.8 Max (2.4T-A95B) and MiMo-V2.6-Pro. Models are ranked on human votes from LMArena and on writing and instruction-following tests. Each model shows the download that fits and the command to run it.

324 models · scores combine 19 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]

Scores use reviewed chat & writing evaluations only. Compare scores within this task. Models without task measurements are excluded.

Filter by your hardware and budget · local only · reset

+ more

set[all sets][LMArena (text) · 579][IFEval (aggregation unspecified) · 3][IFEval (aggregation unspecified) · 2][IFEval (aggregation unspecified) · 2][IFEval prompt-level strict · 2]

Filter the model list while keeping task scores. Models without a shared score scale are not yet calibrated; select a set to see its own scores.

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]

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

detecting your OS…

Score vs memory to run

     │                         Gemma 4 31B        GLM-5.3-Flash ●─●───∘────●───∘──●─∘──∘──────● ∘
     │   ●─────●───●────∘─∘─∘─●∘────∘─∘●∘●●∘─∘∘∘─∘∘∘∘∘∘─∘∘∘∘∘∘∘●∘∘∘∘∘∘∘    ∘   ∘∘       ∘∘ ∘∘  ∘          ∘
  50 ┤ · ∘ · · ∘ · ∘ · ·∘·∘· ·∘∘ ∘ ·∘∘ ∘ ·∘∘ ∘∘∘∘· ∘∘∘ ∘ ∘ ·∘∘∘· Hunyuan Hy3 (295B-A21B) · · ∘ ∘ · · · · · · ·
     │   ∘     ∘   ∘  ∘ ∘ ∘ ∘ ∘∘ ∘   ∘∘   ∘  ∘∘        ∘  ∘    · ∘                             ·
     │                    ∘   ∘       ∘
     └─────────────────────────┬───────────────────────────────┬───────────────────────────────┬──────────────
                             10GB                            100GB                            1TB

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

#	model	task score↓	where↓	RAM↓	context↓	get it	sources
13	Kimi K3 Moonshot AI	74±1*	local	~884GB	1M	llama.cpp ›
20	Qwen3.8 Max (2.4T-A95B) Alibaba	73±1*	local	~380GB	262k	llama.cpp ›
22	MiMo-V2.6-Pro Xiaomi	73±1*	local	~443GB	1M	llama.cpp ›
24	GLM-5.3 Z.ai	73±1*	local	~239GB	1M	llama.cpp ›
27	GLM-5.2 Z.ai	72±1*	local	~121GB	1M	llama.cpp ›
29	DeepSeek V4.1 Flash DeepSeek	72±1*	local	~537GB	1M	how to run ›
30	GLM-5.3-Flash Z.ai	72±1*	local	~103GB	1M	llama.cpp ›
38	MiMo-V2.5-Pro (1T-A42B) Xiaomi	71±1*	local	~163GB	1M	llama.cpp ›
41	DeepSeek V4 Pro (0813) DeepSeek	71±1*	local	~1010GB	1M	llama.cpp ›
42	GLM-5.1 Z.ai	71±1*	local	~119GB	200k	llama.cpp ›
44	Kimi K2.6 Moonshot AI	70±1*	local	~317GB	262k	llama.cpp ›
47	GLM-5 Z.ai	70±1*	local	~119GB	205k	llama.cpp ›
52	Hunyuan Hy3 (295B-A21B) Tencent	70±1*	local	~96GB	256k	llama.cpp ›
54	DeepSeek V4 Pro (Preview, Apr 2026) DeepSeek	69±1*	local	~604GB		llama.cpp ›
57	Gemma 4 31B Google	69±1*	local	~21GB	262k	LM Studio ›
59	MiMo-V2.6-Flash (309B-A15B) Xiaomi	69±1*	local	~136GB	1M	llama.cpp ›
60	Kimi K2.5 Moonshot AI	69±1*	local	~160GB	262k	llama.cpp ›
70	GLM-4.7 Z.ai	68±1*	local	~58GB	205k	llama.cpp ›
71	Inkling (975B-A41B) Thinking Machines	68±1*	local	~156GB	1M	llama.cpp ›
72	Nemotron Cascade 2 30B-A3B NVIDIA	68±4*†	local	~20GB	256k	LM Studio ›
73	Qwen3.5 397B-A17B Alibaba	68±1*	local	~65GB	262k	llama.cpp ›
74	MiniMax-M3 MiniMax	68±1*	local	~137GB	1M	llama.cpp ›
76	DeepSeek V4 Flash (Preview, Apr 2026) DeepSeek	67±1*	local	~103GB		llama.cpp ›
77	Qwen3.8 27B Alibaba	67±1*	local	~18GB	262k	LM Studio ›
78	Gemma 4 26B-A4B Google	67±1*	local	~18GB	262k	LM Studio ›
82	MiMo-V2.5 (310B-A15B) Xiaomi	67±1*	local	~100GB	1M	llama.cpp ›
83	Gemma 4 12B Google	67±6*	local	~8.5GB	262k	LM Studio ›
87	Kimi K2 Thinking Moonshot AI	66±1*	local	~160GB	262k	llama.cpp ›
90	Mistral Medium 3.5 (128B) Mistral	66±1*	local	~42GB	262k	llama.cpp ›
91	Nemotron 3 Ultra 550B-A55B NVIDIA	66±1*	local	~89GB	1M	llama.cpp ›
93	Muse Glimmer 30B Meta	66±1*	local	~19GB	131k	llama.cpp ›
94	DeepSeek V3.2 Exp DeepSeek	66±1*	local	~726GB		how to run ›
95	GLM-4.6 Z.ai	66±1*	local	~58GB	205k	llama.cpp ›
96	DeepSeek V3.2 DeepSeek	65±1*	local	~110GB	128k	llama.cpp ›
98	DeepSeek R1 0528 DeepSeek	65±1*	local	~110GB	164k	llama.cpp ›
99	Qwen3 235B-A22B Instruct 2507 Alibaba	65±1	local	~76GB	262k	llama.cpp ›
100	Kimi K2 Instruct (0905) Moonshot AI	65±1*	local	~298GB		llama.cpp ›
101	Kimi K2 Instruct (0711) Moonshot AI	65±1*	local	~160GB	131k	llama.cpp ›
102	DeepSeek V3.1 Terminus DeepSeek	65±1*	local	~230GB	164k	llama.cpp ›
103	Qwen3.5 122B-A10B Alibaba	65±1*	local	~75GB	262k	LM Studio ›
104	MiniMax-M2.7 MiniMax	64±1*	local	~75GB	205k	llama.cpp ›
107	DeepSeek V3.1 DeepSeek	64±1	local	~110GB	131k	llama.cpp ›
108	Mistral Large 3 (675B-A41B) Mistral	64±1*	local	~108GB	262k	llama.cpp ›
109	Qwen3-VL 235B-A22B Instruct Alibaba	64±1	local	~134GB	256k	LM Studio ›
111	Spark-X2.5 4B iFLYTEK (SparkLLM)	64±6*†	local	~3.9GB	1M	LM Studio ›
113	Qwen3.5 27B Alibaba	64±1*	local	~18GB	262k	LM Studio ›
114	GLM-4.5 Z.ai	64±1	local	~58GB	131k	llama.cpp ›
115	Hy3 preview (295B-A21B) Tencent	64±1*	local	~629GB		how to run ›
118	K-EXAONE 2.0 (750B-A37B) LG AI Research	63±4*†	local	~791GB	262k	how to run ›
119	Inkling Small (266B-A12B) Thinking Machines	63±1*	local	~86GB	1M	llama.cpp ›
123	LongCat-Flash-Chat (560B-A27B) Meituan	63±1*	local	~333GB	128k	LM Studio ›
124	Nemotron-Cascade-8B NVIDIA	63±7*†	local	~6.3GB		LM Studio ›
126	Qwen3 235B-A22B Thinking 2507 Alibaba	63±1	local	~76GB	131k	llama.cpp ›
128	Qwen3-Next 80B-A3B Instruct Alibaba	63±1	local	~27GB	131k	llama.cpp ›
129	DeepSeek R1 (Jan '25) DeepSeek	63±1	local	~110GB	128k	llama.cpp ›
131	LFM2.5 2.6B Liquid AI	62±7*	local	~3.2GB	131k	LM Studio ›
133	DeepSeek V3 0324 DeepSeek	62±1*	local	~108GB	164k	llama.cpp ›
134	Step 3.5 Flash (196B-A11B) StepFun	62±1*	local	~118GB	256k	LM Studio ›
135	Qwen3-VL 235B-A22B Thinking Alibaba	62±1*	local	~111GB	256k	LM Studio ›
136	Qwen3.5 35B-A3B Alibaba	62±1	local	~23GB	262k	LM Studio ›
137	MiniMax-M2.5 MiniMax	61±1*	local	~75GB	205k	llama.cpp ›
141	Qwen3-Coder 480B-A35B Instruct Alibaba	61±1*	local	~78GB	262k	llama.cpp ›
144	MiMo-V2-Flash (309B-A15B) Xiaomi	61±1*	local	~108GB		llama.cpp ›
145	MiniMax-M2.1 MiniMax	61±1*	local	~79GB		llama.cpp ›
146	LFM2.5 8B-A1B Liquid AI	60±5*	local	~6.5GB	128k	LM Studio ›
148	Qwen3 30B-A3B Instruct 2507 Alibaba	60±1	local	~20GB	262k	LM Studio ›
150	GLM-4.6V Z.ai	60±1*	local	~51GB	128k	llama.cpp ›
151	Trinity-Large-Preview Arcee AI	60±1*	local	~155GB		llama.cpp ›
152	LongCat 2.0 (1.6T-A48B) Meituan	60±7*†	local	~2155GB	1M	how to run ›
153	K-EXAONE (236B-A23B) LG AI Research	60±3*†	local	~152GB	262k	llama.cpp ›
154	Spark-X2.5 1.7B iFLYTEK (SparkLLM)	60±5*†	local	~2.7GB	1M	llama.cpp ›
156	Qwen3 235B-A22B (Apr 2025, hybrid) Alibaba	59±1	local	~76GB	131k	llama.cpp ›
157	GLM-4.5-Air Z.ai	59±1*	local	~35GB	131k	llama.cpp ›
159	Gemma 2 9B IT DPO (Princeton) Princeton NLP	59±2*	local	~7.6GB		llama.cpp ›
160	Qwen3-Next 80B-A3B Thinking Alibaba	59±1	local	~27GB	131k	llama.cpp ›
163	EXAONE 4.5 33B LG AI Research	59±4*†	local	~24GB	262k	llama.cpp ›
164	LFM2.5 1.2B Thinking Liquid AI	59±6*†	local	~2.2GB	33k	LM Studio ›
165	Trinity-Large-Thinking Arcee AI	58±1*	local	~141GB	512k	llama.cpp ›
166	GLM-4.7-Flash Z.ai	58±1*	local	~19GB	200k	LM Studio ›
167	MiniMax-M1 (80k) MiniMax	58±1*	local	~959GB	1M	how to run ›
169	Qwen3.5 9B Alibaba	58±3	local	~7.4GB	262k	llama.cpp ›
170	Gemma 3 27B Google	58±1	local	~18GB	131k	LM Studio ›
171	Qwen3 30B-A3B Thinking 2507 Alibaba	58±2	local	~20GB	262k	LM Studio ›
172	Nemotron 3 Super 120B-A12B NVIDIA	58±1*	local	~73GB	262k	LM Studio ›
173	DeepSeek V3 (Dec '24) DeepSeek	57±1	local	~337GB	131k	llama.cpp ›
175	Granite 4.1 30B IBM	57±5*†	local	~19GB		Ollama ›
176	Phi-4-reasoning-plus 14B Microsoft	57±3†	local	~10GB		LM Studio ›
177	Qwen3-VL 32B Thinking Alibaba	57±3*†	local	~22GB	256k	LM Studio ›
178	Mistral Small 3.2 24B Mistral	57±1*	local	~16GB	128k	LM Studio ›
179	INTELLECT-3 Prime Intellect	57±1*	local	~48GB		llama.cpp ›
180	Command A (03-2025) Cohere	57±1*	local	~37GB	256k	llama.cpp ›
182	GLM-4.5V Z.ai	57±1*	local	~69GB	64k	llama.cpp ›
183	gpt-oss-120b OpenAI	56±1*	local	~67GB	131k	llama.cpp ›
185	Llama 3.1 Nemotron Ultra 253B v1 NVIDIA	56±1	local	~82GB	128k	llama.cpp ›
186	Nemotron 3.5 Lightning 30B-A3B NVIDIA	56±1*	local	~20GB	262k	LM Studio ›
187	Qwen3.5 4B Alibaba	56±3	local	~4.3GB	262k	llama.cpp ›
188	Nemotron 3 Nano 4B NVIDIA	56±6*	local	~3.8GB	262k	LM Studio ›
189	LFM2.5 1.2B Instruct Liquid AI	56±6*†	local	~2.2GB	33k	LM Studio ›
190	Qwen3 32B Alibaba	56±1	local	~20GB	131k	LM Studio ›
192	QwQ-32B Alibaba	56±1	local	~21GB	131k	LM Studio ›
193	Mellum2.1 12B-A2.5B Thinking JetBrains	56±8*†	local	~8.3GB		LM Studio ›
194	MiniMax-M2 MiniMax	55±1*	local	~75GB	205k	llama.cpp ›
195	Llama 3.3 Nemotron Super 49B v1.5 NVIDIA	55±1*	local	~18GB	131k	llama.cpp ›
198	Gemma 3 12B Google	55±1*	local	~9.1GB	131k	LM Studio ›
199	Granite 4.2 30B IBM	55±1*	local	~19GB	131k	LM Studio ›
201	Qwen3-VL 30B-A3B Instruct Alibaba	55±3*†	local	~21GB	256k	LM Studio ›
202	MiniCPM5 2B OpenBMB	55±6*†	local	~3GB	131k	LM Studio ›
205	Phi-4-reasoning 14B Microsoft	54±3†	local	~10GB	33k	LM Studio ›
208	Qwen3 8B Alibaba	54±3*†	local	~6.1GB	33k	LM Studio ›
209	Qwen3 4B Instruct 2507 Alibaba	54±2†	local	~3.9GB	262k	LM Studio ›
210	Qwen3-VL 32B Instruct Alibaba	54±3*†	local	~22GB	256k	LM Studio ›
211	Llama 3.3 Nemotron Super 49B v1 NVIDIA	54±1	local	~22GB		llama.cpp ›
212	Nemotron Nano 9B v2 NVIDIA	54±3	local	~7.6GB	131k	llama.cpp ›
214	Olmo 3.1 32B Instruct Ai2	54±1*	local	~21GB	66k	LM Studio ›
215	Qwen3 4B Thinking 2507 Alibaba	53±3*†	local	~3.9GB	262k	LM Studio ›
216	Llama 3.1 405B Instruct Meta	53±1	local	~160GB	128k	llama.cpp ›
219	Nemotron-Cascade-14B-Thinking NVIDIA	53±6*†	local	~10GB		LM Studio ›
220	Qwen3-VL 8B Instruct Alibaba	53±3*†	local	~6.8GB	256k	llama.cpp ›
222	Qwen3 30B-A3B (Apr 2025, hybrid) Alibaba	53±1	local	~19GB	33k	LM Studio ›
223	Llama 4 Maverick (402B-A17B) Meta	53±1	local	~65GB	1M	llama.cpp ›
224	Seed-OSS 36B Instruct ByteDance Seed	53±5*†	local	~23GB	524k	LM Studio ›
225	DeepSeek-V2.5 (1210) DeepSeek	53±1*	local	~82GB		llama.cpp ›
226	Llama 4 Scout (109B-A17B) Meta	53±1*	local	~36GB	328k	llama.cpp ›
227	EXAONE 4.0 32B LG AI Research	53±4*†	local	~20GB	131k	LM Studio ›
228	Molmo 2 8B Ai2	53±1*	local	~6.8GB		llama.cpp ›
229	Qwen3-VL 8B Thinking Alibaba	53±3*†	local	~6.8GB	256k	llama.cpp ›
230	Olmo 3 7B Think Ai2	53±3	local	~5.8GB	66k	LM Studio ›
231	gpt-oss-20b OpenAI	52±1*	local	~14GB	131k	llama.cpp ›
233	Athene-V2-Chat 72B Nexusflow	52±1	local	~44GB	33k	LM Studio ›
234	Qwen3-VL 4B Thinking Alibaba	52±3*†	local	~4.1GB	256k	llama.cpp ›
235	Gemma 3n E4B Google	52±1*	local	~6.2GB	32k	llama.cpp ›
238	Qwen3-VL 4B Instruct Alibaba	52±3*†	local	~4.1GB	256k	llama.cpp ›
239	Mistral Large 2 (2407) Mistral	52±1	local	~45GB	128k	llama.cpp ›
240	Granite 4.0 H Micro 3B IBM	52±5*†	local	~3.4GB	131k	LM Studio ›
242	Llama 3.3 70B Instruct Meta	52±1	local	~24GB	128k	llama.cpp ›
243	Nemotron 3 Nano 30B-A3B NVIDIA	52±1*	local	~20GB	262k	LM Studio ›
244	Qwen3-VL 30B-A3B Thinking Alibaba	51±3*†	local	~21GB	256k	LM Studio ›
245	Athene-70B Nexusflow	51±1*	local	~29GB		llama.cpp ›
246	Olmo 3 32B Think Ai2	51±1	local	~21GB	66k	LM Studio ›
250	Qwen2.5 72B Instruct Alibaba	51±1	local	~30GB	33k	llama.cpp ›
251	Llama 3.1 Nemotron 70B Instruct NVIDIA	51±1	local	~43GB	128k	LM Studio ›
252	DeepSeek-V2.5 DeepSeek	51±1*	local	~151GB	164k	Ollama ›
253	Granite 4.1 8B IBM	51±1*	local	~7GB	131k	LM Studio ›
254	Gemma 2 9B IT SimPO Princeton NLP	51±1*	local	~7.6GB		llama.cpp ›
255	Mistral Large 2 (2411) Mistral	51±1*	local	~78GB	128k	llama.cpp ›
256	Olmo 3 7B Instruct Ai2	50±3	local	~5.8GB	66k	LM Studio ›
258	Gemma 3 4B Google	50±1	local	~3.8GB	131k	LM Studio ›
259	Granite 4.1 3B IBM	50±5*†	local	~3.7GB		Ollama ›
260	Mistral Small 3.1 24B (2503) Mistral	50±1*	local	~17GB	128k	llama.cpp ›
261	Granite 4.0 Micro 3B IBM	50±5*†	local	~3.5GB	131k	LM Studio ›
262	Command R7B (12-2024) Cohere	49±3*	local	~6.2GB	128k	LM Studio ›
263	Flan-T5 XXL Google	49±15*	local	~49GB		how to run ›
265	Qwen2.5 32B Instruct Alibaba	49±2	local	~21GB	131k	LM Studio ›
266	Llama-3-Instruct-8B-SimPO-ExPO Chujie Zheng (ExPO)	49±2*	local	~6.6GB		llama.cpp ›
267	Granite 4.2 3B IBM	49±1	local	~3.7GB	131k	LM Studio ›
268	Granite 4.2 8B IBM	49±1*	local	~6.6GB	131k	LM Studio ›
269	Olmo 3.1 32B Think Ai2	49±1	local	~21GB	66k	LM Studio ›
270	Llama 3.1 Tulu 3 70B DPO Ai2	49±4*†	local	~22GB	131k	llama.cpp ›
271	Granite 4.0 H Small (32B-A9B) IBM	49±1	local	~21GB	131k	LM Studio ›
273	Jamba 1.5 Large AI21 Labs	49±1	local	~839GB		how to run ›
274	Llama-3-Instruct-8B-SimPO Princeton NLP	49±2*	local	~6.6GB		llama.cpp ›
275	Granite 4.0 H Tiny (7B-A1B) IBM	49±4*†	local	~5.6GB	131k	LM Studio ›
276	Gemma 2 27B IT Google	48±1	local	~18GB	8k	LM Studio ›
277	Granite 4.0 H 1B IBM	48±4	local	~2.3GB	131k	LM Studio ›
278	Llama 3.1 70B Instruct Meta	48±1	local	~27GB	128k	llama.cpp ›
280	Llama 3.1 Tulu 3 70B Ai2	48±1	local	~43GB	131k	LM Studio ›
281	Llama 3.1 Nemotron 51B Instruct NVIDIA	48±1*	local	~34GB		llama.cpp ›
282	s1.1 32B Simple Scaling	48±4*	local	~22GB	33k	LM Studio ›
283	Mellum2 12B-A2.5B Instruct JetBrains	48±4*†	local	~8.6GB	131k	LM Studio ›
285	LFM2 2.6B Liquid AI	48±6*†	local	~3GB	33k	LM Studio ›
286	Qwen3 4B Alibaba	47±2	local	~3.7GB	33k	LM Studio ›
288	Granite 4.0 1B IBM	47±3	local	~2.4GB	128k	LM Studio ›
289	OLMo 2 13B Instruct (1124) Allen Institute for AI	47±3*†	local	~9.6GB	4k	LM Studio ›
290	Qwen3.5 2B Alibaba	47±3	local	~2.9GB	262k	llama.cpp ›
291	Qwen2.5 14B Instruct Alibaba	47±3	local	~10GB	131k	LM Studio ›
295	Command R Plus (08-2024) Cohere	47±1*	local	~63GB	128k	LM Studio ›
296	Mellum2 12B-A2.5B Thinking JetBrains	47±3*†	local	~8.6GB	131k	LM Studio ›
298	Mistral Small 3 24B (2501) Mistral	47±1*	local	~16GB	32k	LM Studio ›
299	Qwen3-VL 2B Thinking Alibaba	47±3*†	local	~2.7GB	256k	llama.cpp ›
302	Qwen2.5 Coder 32B Instruct Alibaba	46±1*	local	~21GB	131k	LM Studio ›
303	Aya Expanse 32B Cohere	46±1*	local	~22GB	128k	llama.cpp ›
304	North Micro Vision Instruct (2.4B) Cohere	46±8*†	local	~3.8GB	128k	how to run ›
305	Mistral Small 4 (119B-A6.5B) Mistral	46±7*†	local	~40GB	262k	llama.cpp ›
306	Gemma 2 9B IT Google	46±1	local	~7GB	8k	LM Studio ›
307	Llama 3 70B Instruct Meta	46±1	local	~24GB	8k	llama.cpp ›
308	DeepSeek-Coder-V2 Instruct (0614) DeepSeek	46±1	local	~92GB		llama.cpp ›
309	Qwen2.5 7B Instruct Alibaba	45±2	local	~6GB	131k	LM Studio ›
311	OpenThinker2 32B Open Thoughts	45±5*	local	~21GB	33k	LM Studio ›
313	Qwen2 72B Instruct Alibaba	45±1	local	~33GB	131k	llama.cpp ›
315	Granite 3.3 8B Instruct IBM	45±4*†	local	~6.3GB	128k	LM Studio ›
316	LFM2 1.2B Liquid AI	45±4*†	local	~2.2GB	33k	LM Studio ›
318	Command R Plus (04-2024) Cohere	44±1	local	~43GB	128k	llama.cpp ›
319	Phi-4 14B Microsoft	44±1	local	~10GB	16k	LM Studio ›
320	OLMo 2 32B Instruct (0325) Ai2	44±1	local	~21GB		LM Studio ›
321	Granite 3.2 8B Instruct IBM	44±4*†	local	~6.6GB	131k	llama.cpp ›
322	EXAONE 3.5 2.4B Instruct LG AI Research	44±3*†	local	~3GB	33k	LM Studio ›
323	Qwen3 1.7B Alibaba	44±2	local	~2.6GB	33k	LM Studio ›
324	Qwen2.5 3B Instruct Alibaba	44±3*	local	~3.3GB	33k	LM Studio ›
325	Qwen2 7B Instruct Alibaba	44±3*	local	~6GB	131k	LM Studio ›
326	Command R (08-2024) Cohere	44±1*	local	~21GB	128k	LM Studio ›
327	Ministral 3 14B Mistral	43±3*	local	~9.1GB	262k	LM Studio ›
328	LFM2 700M Liquid AI	43±4*†	local	~1.9GB	33k	LM Studio ›
329	Qwen1.5 110B Chat Alibaba	43±1	local	~45GB	32k	llama.cpp ›
330	Qwen3-VL 2B Instruct Alibaba	43±2*	local	~2.7GB	256k	llama.cpp ›
333	Jamba 1.5 Mini AI21 Labs	42±1	local	~34GB		llama.cpp ›
334	SmolLM3 3B Hugging Face	42±4*	local	~3.3GB	66k	LM Studio ›
335	EXAONE 4.0 1.2B LG AI Research	42±3*†	local	~2.2GB	66k	LM Studio ›
337	Ministral 8B (2410) Mistral	42±1	local	~6.2GB	128k	LM Studio ›
338	Ministral 3 8B Mistral	42±6*†	local	~6.5GB	262k	LM Studio ›
339	OLMo 2 7B Instruct (1124) Allen Institute for AI	42±4*†	local	~5.8GB	4k	LM Studio ›
340	Qwen1.5 72B Chat Alibaba	41±1	local	~31GB	32k	llama.cpp ›
341	Mixtral 8x22B Instruct v0.1 Mistral	41±1	local	~56GB	64k	llama.cpp ›
342	Aya Expanse 8B Cohere	41±1*	local	~6.2GB	8k	LM Studio ›
345	Llama 3.1 Tulu 3 8B Ai2	40±1	local	~6.2GB	131k	LM Studio ›
347	Command R (03-2024) Cohere	40±1	local	~24GB	128k	llama.cpp ›
350	Xwin-LM 70B v0.1 Xwin-LM	39±2*	local	~28GB		llama.cpp ›
351	Llama 3 8B Instruct Meta	39±1	local	~6.6GB	8k	llama.cpp ›
353	LFM2 350M Liquid AI	39±4*†	local	~1.7GB	33k	LM Studio ›
354	Zephyr ORPO 141B-A35B v0.1 Hugging Face	39±1	local	~56GB	66k	llama.cpp ›
355	Llama 3.1 8B Instruct Meta	39±1	local	~6.2GB	128k	LM Studio ›
356	Granite 3.1 8B Instruct IBM	39±1	local	~6.6GB	128k	llama.cpp ›
357	Yi 1.5 34B Chat 01.AI	39±1*	local	~22GB	4k	LM Studio ›
359	Gemma 3 1B Google	38±2	local	~2.2GB	32k	LM Studio ›
360	Qwen3 0.6B Alibaba	38±3*	local	~1.8GB	33k	LM Studio ›
361	Ministral 3 14B Reasoning 2512 Mistral	38±3*	local	~10GB	262k	llama.cpp ›
362	Qwen1.5 32B Chat Alibaba	38±1*	local	~22GB	32k	llama.cpp ›
363	Phi-3-medium-4k-instruct Microsoft	38±1	local	~9.8GB	4k	LM Studio ›
364	Gemma 2 2B IT Google	37±1*	local	~3.1GB	8k	LM Studio ›
365	Phi-3.5-vision-instruct Microsoft	37±8*	local	~10GB		how to run ›
366	Mixtral 8x7B Instruct v0.1 Mistral	37±1	local	~29GB	32k	LM Studio ›
368	DBRX Instruct Databricks	37±1	local	~85GB	32k	Ollama ›
369	InternLM2.5 20B Chat Shanghai AI Laboratory (InternLM)	37±1*	local	~14GB	33k	llama.cpp ›
370	InternLM2 Chat 20B Shanghai AI Laboratory (InternLM)	36±2*	local	~14GB		llama.cpp ›
371	Mistral 7B Instruct v0.3 Mistral	36±2*	local	~5.8GB	32k	LM Studio ›
372	Yi 34B Chat 01.AI	36±1	local	~23GB	4k	llama.cpp ›
373	Qwen1.5 14B Chat Alibaba	36±1	local	~11GB	32k	Ollama ›
374	Granite 3.3 2B Instruct IBM	36±3*†	local	~3GB	128k	LM Studio ›
375	WizardLM 70B V1.0 Microsoft	35±1	local	~32GB		llama.cpp ›
376	OpenChat 3.5 0106 OpenChat	35±1*	local	~6.1GB	8k	llama.cpp ›
377	Tulu 2 DPO 70B Ai2	35±1	local	~32GB		llama.cpp ›
378	Granite 3.1 3B-A800M Instruct IBM	35±5*†	local	~3.6GB	131k	llama.cpp ›
379	Snowflake Arctic Instruct Snowflake	35±1	local	~1007GB		how to run ›
380	Granite 3.0 8B Instruct IBM	35±1*	local	~6.6GB	4k	llama.cpp ›
381	OpenChat 3.5 OpenChat	35±1*	local	~6.1GB	8k	llama.cpp ›
382	DeepSeek LLM 67B Chat DeepSeek	35±1	local	~31GB	4k	llama.cpp ›
384	Gemma 4 E4B Google	35±2	local	~6.9GB	131k	LM Studio ›
385	Gemma 1.1 7B IT Google	35±1	local	~7.1GB	8k	llama.cpp ›
386	Granite 3.1 2B Instruct IBM	35±1	local	~3.1GB	128k	llama.cpp ›
387	Phi-3-small-8k-instruct Microsoft	34±1	local	~17GB		how to run ›
388	Starling-LM 7B beta Nexusflow	34±1	local	~6GB	8k	LM Studio ›
389	Ministral 3 3B Mistral	34±5*	local	~3.5GB	262k	LM Studio ›
390	Deita 7B v1.0 HKUST NLP	34±2*	local	~32GB		how to run ›
391	Granite 4.0 H 350M IBM	34±4*†	local	~1.7GB	33k	LM Studio ›
392	Vicuna 33B v1.3 LMSYS	34±1	local	~23GB	2k	Ollama ›
393	OpenHermes 2.5 Mistral 7B Nous Research	34±1	local	~6.1GB		llama.cpp ›
394	Granite 3.2 2B Instruct IBM	34±3*†	local	~3GB	131k	LM Studio ›
395	Llama 3.2 3B Instruct Meta	33±1	local	~3.4GB	131k	LM Studio ›
396	Llama 2 70B Chat Meta	33±1	local	~32GB	4k	llama.cpp ›
397	OLMo 2 1B Instruct (0425) Allen Institute for AI	33±4*†	local	~2.4GB	4k	llama.cpp ›
398	Starling-LM 7B alpha UC Berkeley (NEST)	33±1	local	~6.1GB	8k	llama.cpp ›
399	OpenThinker3 7B Open Thoughts	33±3*	local	~6GB	33k	LM Studio ›
400	Nous Hermes 2 Mixtral 8x7B DPO Nous Research	33±1*	local	~31GB	33k	llama.cpp ›
401	Gemma 3 270M IT Google	32±5*†	local	~1.7GB	32k	LM Studio ›
402	Granite 3.0 2B Instruct IBM	32±1*	local	~3.2GB	4k	llama.cpp ›
403	SmolLM2 360M Instruct Hugging Face	32±5*†	local	~1.7GB	8k	LM Studio ›
404	QwQ-32B-Preview Alibaba	31±1*	local	~21GB	33k	LM Studio ›
405	WizardLM 13B V1.2 Microsoft	31±1	local	~9.8GB		llama.cpp ›
406	Mistral 7B Instruct v0.2 Mistral	31±1	local	~6GB	32k	LM Studio ›
407	Tulu 2 DPO 13B Allen Institute for AI	31±2*	local	~9.8GB		llama.cpp ›
408	Granite 3.1 1B-A400M Instruct IBM	31±5*†	local	~2.3GB	131k	llama.cpp ›
409	Granite 3.0 3B-A800M Instruct IBM	31±4*†	local	~3.7GB	4k	llama.cpp ›
410	SOLAR 10.7B Instruct v1.0 Upstage	31±1*	local	~8.3GB	4k	llama.cpp ›
411	Qwen3.5 0.8B Alibaba	31±4*†	local	~2GB	262k	llama.cpp ›
412	Dolphin 2.2.1 Mistral 7B Cognitive Computations	31±1	local	~6.1GB		llama.cpp ›
413	Qwen1.5 7B Chat Alibaba	31±1	local	~6.5GB	32k	llama.cpp ›
414	Falcon-180B-Chat TII	30±1*	local	~115GB	2k	Ollama ›
415	Qwen2.5 0.5B Instruct Alibaba	30±3*	local	~1.8GB	33k	LM Studio ›
417	Qwen-14B-Chat Alibaba	30±1	local	~11GB		llama.cpp ›
418	Phi-3-mini-4k-instruct (June 2024 update) Microsoft	30±1*	local	~3.7GB		LM Studio ›
419	Vicuna 13B v1.5 LMSYS	30±1	local	~9.8GB	4k	llama.cpp ›
420	Granite 4.0 350M IBM	30±4*†	local	~1.7GB	32k	LM Studio ›
421	Llama 2 13B Chat Meta	29±1	local	~9.8GB	4k	llama.cpp ›
422	Code Llama 34B Instruct Meta	29±1*	local	~22GB	16k	LM Studio ›
423	Zephyr 7B beta Hugging Face	29±1	local	~6.1GB		llama.cpp ›
424	Tulu 2 DPO 7B Allen Institute for AI	29±2*	local	~5.8GB		llama.cpp ›
425	Phi-3-mini-128k-instruct (Apr 2024) Microsoft	29±1	local	~9.5GB		how to run ›
426	Gemma 7B IT Google	29±1	local	~7.1GB	8k	llama.cpp ›
427	Phi-3-mini-4k-instruct (Apr 2024) Microsoft	29±1*	local	~4GB		llama.cpp ›
428	Zephyr 7B alpha Hugging Face	28±1	local	~6.1GB		llama.cpp ›
430	Gemma 4 E2B Google	28±3	local	~5.1GB	131k	llama.cpp ›
431	Llama 3.2 1B Instruct Meta	27±1	local	~2.2GB	131k	LM Studio ›	+1
432	Code Llama 70B Instruct Meta	27±1*	local	~28GB	16k	llama.cpp ›
433	Guanaco 33B University of Washington	27±1	local	~70GB		how to run ›
434	Vicuna 7B v1.5 LMSYS	27±1	local	~5.8GB	4k	llama.cpp ›
435	Gemma 1.1 2B IT Google	27±1	local	~3.2GB	8k	llama.cpp ›
436	SmolLM2 1.7B Instruct Hugging Face	27±1	local	~2.6GB	8k	LM Studio ›
437	Granite 3.0 1B-A400M Instruct IBM	26±4*†	local	~2.3GB	4k	llama.cpp ›
438	Qwen2.5 1.5B Instruct Alibaba	26±3	local	~2.4GB	33k	LM Studio ›
439	Mistral 7B Instruct v0.1 Mistral	26±1*	local	~6.1GB	8k	llama.cpp ›
440	Llama 2 7B Chat Meta	25±1	local	~5.8GB	4k	llama.cpp ›
441	Ghost 7B Alpha Ghost X	25±2*	local	~6.1GB		llama.cpp ›
442	Nous-Hermes-13b Nous Research	25±2*	local	~9.8GB	2k	llama.cpp ›
443	Qwen1.5 4B Chat Alibaba	24±1*	local	~4.1GB	32k	llama.cpp ›
444	Gemma 2B IT Google	23±1	local	~3.1GB	8k	llama.cpp ›
445	OLMo 7B Instruct Ai2	21±1*	local	~16GB		how to run ›
448	GPT4All-13B-Snoozy Nomic AI	20±1	local	~56GB		how to run ›
450	Falcon 40B Instruct TII	19±1	local	~27GB		Ollama ›
451	ChatGLM3-6B Zhipu AI (THUDM)	19±1*	local	~5.8GB		llama.cpp ›
452	Falcon 7B Instruct TII	18±2*	local	~5.9GB		Ollama ›
454	ChatGLM2-6B Zhipu AI (THUDM)	16±1	local	~15GB		how to run ›
455	Qwen1.5 1.8B Chat Alibaba	16±5*	local	~2.7GB	33k	Ollama ›
456	Phi-2 (2.7B) Microsoft	14±2*	local	~3.2GB	2k	LM Studio ›
457	OpenAssistant SFT-4 Pythia 12B LAION (OpenAssistant)	14±1	local	~9.5GB		llama.cpp ›
459	ChatGLM-6B Zhipu AI (THUDM)	12±1	local	~16GB		how to run ›
460	FastChat-T5 3B LMSYS	10±1	local	~8.5GB		how to run ›
463	StableLM-Tuned-Alpha 7B Stability AI	6±1	local	~6.9GB		llama.cpp ›
—	ERNIE 4.5 300B-A47B Baidu	not yet calibrated	local	~97GB	131k	llama.cpp ›
—	Hermes 3 Llama 3.1 405B Nous Research	not yet calibrated	local	~114GB	131k	llama.cpp ›
—	Hermes 3 Llama 3.1 70B Nous Research	not yet calibrated	local	~44GB	131k	Ollama ›
—	Hermes 3 Llama 3.1 8B Nous Research	not yet calibrated	local	~6.6GB	128k	llama.cpp ›
—	Phi-3.5-mini-instruct Microsoft	not yet calibrated	local	~3.7GB	128k	LM Studio ›
—	Phi-4-mini 3.8B Microsoft	not yet calibrated	local	~3.8GB	128k	LM Studio ›
—	Qwen2 0.5B Instruct Alibaba	not yet calibrated	local	~1.8GB	32k	LM Studio ›
—	Qwen2 1.5B Instruct Alibaba	not yet calibrated	local	~2.4GB	32k	LM Studio ›

+ 148 more don't fit your machine or budget · show them

+ 71 models with fewer than 3 results left off, since one or two tests say little about overall ability · show them

472 of 588 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