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

Best AI models for agents to run locally

A leaderboard of the best AI models for agents you can run on your own computer. DeepSeek V4.1 Flash is #1 right now, followed by GLM-5.3 and MiMo-V2.6-Pro. Models are ranked on agent tests like SWE-bench and Terminal-Bench, where they use tools, write code and work in a terminal. Each model shows the download that fits and the command to run it.

233 models · scores combine 76 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 agent tasks 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][BFCL v3 · 374][BFCL v3 · 4][SWE-bench Multilingual · 4][SWE-bench Verified · 2][SWE-bench Multilingual · 2][BFCL V4 (December 2025 archive) · 2][τ²-bench Telecom · 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][$50][any]max per M

detecting your OS…

Score vs memory to run

     │                                            GLM-5.3-Flash ●──GLM-5.3─●∘─────∘─∘──●      ∘
  75 ┤ · · · · ●───────────Qwen3.8 27B─●∘─∘∘───────∘────●∘─∘∘∘∘∘∘∘∘·∘·∘· · · · ∘∘· · ·∘·∘∘ ·∘· ·∘· · · · ·∘· ·
  50 ┤ · ●─────∘ · ∘ · ·∘·∘·∘·∘∘ · ·∘·∘∘∘∘∘∘ ∘ ∘∘∘∘∘∘· ·∘·∘∘∘∘ ·∘∘ ∘ ∘∘· · ∘ · ·∘· · ∘ · · ∘ ∘ · · · · · · · ·
  25 ┤ · ∘ · · ∘ · ∘ ·∘·∘·∘·∘·∘∘ · · ∘ ∘ ·∘∘ ∘∘· ∘ ·∘· ∘ ∘ · ∘ · · · ∘ · · · · · · · · · · · · · · · · · · · ·
     │   ∘                       ∘        ∘
     └─────────────────────────┬───────────────────────────────┬───────────────────────────────┬──────────────
                             10GB                            100GB                            1TB

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

#	model	task score↓	where↓	RAM↓	context↓	get it	sources
12	DeepSeek V4.1 Flash DeepSeek	79±1	local	~537GB	1M	how to run ›
14	GLM-5.3 Z.ai	79±1*	local	~239GB	1M	llama.cpp ›
17	MiMo-V2.6-Pro Xiaomi	78±2	local	~443GB	1M	llama.cpp ›
18	Qwen3.8 Max (2.4T-A95B) Alibaba	77±2	local	~380GB	262k	llama.cpp ›
21	Hy4 preview (770B-A49B) Tencent	77±2	local	~248GB	1M	llama.cpp ›
23	Kimi K3 Moonshot AI	76±1	local	~884GB	1M	llama.cpp ›
24	GLM-5.3-Flash Z.ai	76±1	local	~103GB	1M	llama.cpp ›
28	MiMo-V2.6-Flash (309B-A15B) Xiaomi	75±2	local	~136GB	1M	llama.cpp ›
31	DeepSeek V4 Pro (0813) DeepSeek	74±1	local	~1010GB	1M	llama.cpp ›
32	DeepSeek V4 Flash Vision (Exp) DeepSeek	74±3†	local	~104GB	1M	llama.cpp ›
40	GLM-5.2 Z.ai	73±1	local	~121GB	1M	llama.cpp ›	+3
41	Qwen3.8-Flash-Next Alibaba	72±2	local	~59GB	262k	llama.cpp ›
44	DeepSeek V4 Flash (0731) DeepSeek	71±2	local	~92GB	1M	llama.cpp ›
46	Qwen3.8 27B Alibaba	71±1	local	~18GB	262k	LM Studio ›
49	Hunyuan Hy3 (295B-A21B) Tencent	70±1	local	~96GB	256k	llama.cpp ›
53	Kimi K2.6 Moonshot AI	68±1	local	~317GB	262k	llama.cpp ›	+2
56	GLM-5.1 Z.ai	67±1	local	~119GB	200k	llama.cpp ›	+5
57	DeepSeek V4 Pro (Preview, Apr 2026) DeepSeek	67±1	local	~604GB		llama.cpp ›
58	MiniMax-M3 MiniMax	67±2	local	~137GB	1M	llama.cpp ›
65	Solar Open 2 (250B-A15B) Upstage	65±2†	local	~527GB	1M	how to run ›
66	MiMo-V2.5 (310B-A15B) Xiaomi	65±2	local	~100GB	1M	llama.cpp ›
67	DeepSeek V4 Flash (Preview, Apr 2026) DeepSeek	65±1	local	~103GB		llama.cpp ›
68	Qwen3.5 397B-A17B Alibaba	65±1	local	~65GB	262k	llama.cpp ›
69	Ornith 1.5 35B-A3B DeepReinforce	65±2†	local	~22GB	262k	LM Studio ›
72	Muse Glimmer 30B Meta	65±2	local	~19GB	131k	llama.cpp ›
74	LongCat 2.0 (1.6T-A48B) Meituan	64±6*	local	~2155GB	1M	how to run ›
75	GLM-5 Z.ai	64±1	local	~119GB	205k	llama.cpp ›	+1
76	Qwen3.6 27B Alibaba	64±1	local	~18GB	262k	LM Studio ›
78	MiMo-V2.5-Pro (1T-A42B) Xiaomi	64±1	local	~163GB	1M	llama.cpp ›
80	Inkling Small (266B-A12B) Thinking Machines	64±2	local	~86GB	1M	llama.cpp ›
81	MiniMax-M2.7 MiniMax	64±2	local	~75GB	205k	llama.cpp ›
83	Inkling (975B-A41B) Thinking Machines	63±2	local	~156GB	1M	llama.cpp ›
84	Step 3.5 Flash (196B-A11B) StepFun	63±2†	local	~118GB	256k	LM Studio ›
86	Hy3 preview (295B-A21B) Tencent	63±2†	local	~629GB		how to run ›
87	Nemotron 3 Ultra 550B-A55B NVIDIA	63±2†	local	~89GB	1M	llama.cpp ›
88	KAT-Coder-V2.5-Dev (35B-A3B) Kuaishou (Kwaipilot)	62±2†	local	~22GB	262k	LM Studio ›
89	Qwen3.5 27B Alibaba	62±1	local	~18GB	262k	LM Studio ›
90	Qwen3.5 122B-A10B Alibaba	62±2	local	~75GB	262k	LM Studio ›
92	Kimi K2.5 Moonshot AI	62±1	local	~160GB	262k	llama.cpp ›	+2
93	Qwen3.6 35B-A3B Alibaba	61±1	local	~23GB	262k	LM Studio ›
95	GLM-4.7 Z.ai	61±1†	local	~58GB	205k	llama.cpp ›
98	Ornith 1.0 35B (35B-A3B) DeepReinforce	60±1	local	~23GB	262k	LM Studio ›
100	Qwen3.5 35B-A3B Alibaba	60±1	local	~23GB	262k	LM Studio ›
101	MiniMax-M2.1 MiniMax	59±2	local	~79GB		llama.cpp ›
102	K-EXAONE 2.0 (750B-A37B) LG AI Research	59±4†	local	~791GB	262k	how to run ›
103	SmolLM3 3B Hugging Face	59±11*†	local	~3.3GB	66k	LM Studio ›
105	DeepSeek V3.2 DeepSeek	59±2	local	~110GB	128k	llama.cpp ›
106	MiniMax-M2.5 MiniMax	59±2	local	~75GB	205k	llama.cpp ›
107	MiMo-V2-Flash (309B-A15B) Xiaomi	58±3*	local	~108GB		llama.cpp ›
108	LongCat-Flash-Chat (560B-A27B) Meituan	58±2†	local	~333GB	128k	LM Studio ›
109	Mistral Medium 3.5 (128B) Mistral	58±2	local	~42GB	262k	llama.cpp ›
110	MiniMax-M2 MiniMax	58±2	local	~75GB	205k	llama.cpp ›
114	Gemma 4 31B Google	57±1	local	~21GB	262k	LM Studio ›	+4
116	GLM-4.6 Z.ai	57±1	local	~58GB	205k	llama.cpp ›
118	Jamba 1.5 Large AI21 Labs	57±10*†	local	~839GB		how to run ›
119	Spark-X2.5 4B iFLYTEK (SparkLLM)	56±2†	local	~3.9GB	1M	LM Studio ›
122	Kimi K2 Thinking Moonshot AI	56±2	local	~160GB	262k	llama.cpp ›
123	EXAONE 4.5 33B LG AI Research	56±3*†	local	~24GB	262k	llama.cpp ›
124	Kimi K2 Instruct (0711) Moonshot AI	56±1	local	~160GB	131k	llama.cpp ›
127	MiMo-V2.6-Distill-Qwen-9B Xiaomi	56±3†	local	~7.8GB		LM Studio ›
132	K-EXAONE (236B-A23B) LG AI Research	55±2	local	~152GB	262k	llama.cpp ›
133	DeepSeek V3.1 Terminus DeepSeek	55±3†	local	~230GB	164k	llama.cpp ›
135	GigaChat 3.5 Ultra Reasoning (432B-A28B) Sber	55±3†	local	~461GB	262k	how to run ›
136	GLM-4.5 Z.ai	55±2	local	~58GB	131k	llama.cpp ›
137	MiniCPM5 2B OpenBMB	55±3†	local	~3GB	131k	LM Studio ›
139	Nemotron 3 Super 120B-A12B NVIDIA	54±2†	local	~73GB	262k	LM Studio ›
142	Granite 4.2 30B IBM	54±2†	local	~19GB	131k	LM Studio ›
144	Qwen3-Coder 480B-A35B Instruct Alibaba	53±2	local	~78GB	262k	llama.cpp ›
148	Nemotron-Labs-3 Puzzle 75B-A9B NVIDIA	53±2†	local	~35GB		llama.cpp ›
149	Seed-OSS 36B Instruct ByteDance Seed	52±3†	local	~23GB	524k	LM Studio ›
151	Mistral Small 4 (119B-A6.5B) Mistral	52±3*	local	~40GB	262k	llama.cpp ›
153	Command A+ (05-2026) Cohere	52±2	local	~71GB	128k	llama.cpp ›
154	DeepSeek V3.2 Exp DeepSeek	52±3*	local	~726GB		how to run ›
155	GLM-4.5-Air Z.ai	51±2	local	~35GB	131k	llama.cpp ›
156	Jamba 1.5 Mini AI21 Labs	51±10*†	local	~34GB		llama.cpp ›
157	Qwen3.5 9B Alibaba	51±1	local	~7.4GB	262k	llama.cpp ›	+3
158	DeepSeek V3.1 DeepSeek	51±1	local	~110GB	131k	llama.cpp ›
159	Mellum2.1 12B-A2.5B Thinking JetBrains	50±3†	local	~8.3GB		LM Studio ›
160	Qwen3 235B-A22B Thinking 2507 Alibaba	50±1	local	~76GB	131k	llama.cpp ›
161	gpt-oss-120b OpenAI	49±2	local	~67GB	131k	llama.cpp ›	+1
162	Qwen3-Next 80B-A3B Thinking Alibaba	49±2†	local	~27GB	131k	llama.cpp ›
163	Command A Reasoning (08-2025) Cohere	49±3	local	~34GB		llama.cpp ›
164	Granite 4.1 30B IBM	49±5*†	local	~19GB		Ollama ›
166	Gemma 4 12B Google	48±2	local	~8.5GB	262k	LM Studio ›
167	Qwen3-VL 235B-A22B Thinking Alibaba	48±2†	local	~111GB	256k	LM Studio ›
168	Gemma 4 26B-A4B Google	48±1	local	~18GB	262k	LM Studio ›	+1
170	Qwen3-VL 32B Thinking Alibaba	48±1	local	~22GB	256k	LM Studio ›
171	Granite 4.2 8B IBM	48±2†	local	~6.6GB	131k	LM Studio ›
174	Qwen3.5 4B Alibaba	47±1	local	~4.3GB	262k	llama.cpp ›	+1
175	Qwen3 235B-A22B Instruct 2507 Alibaba	47±1	local	~76GB	262k	llama.cpp ›	+1
177	Qwen3-VL 32B Instruct Alibaba	46±3*†	local	~22GB	256k	LM Studio ›
180	Solar Open 100B (102B-A12B) Upstage	46±3†	local	~41GB	131k	llama.cpp ›
182	LFM2.5 2.6B Liquid AI	46±2	local	~3.2GB	131k	LM Studio ›
183	GigaChat 3.5 Ultra Instruct (432B-A28B) Sber	46±3†	local	~457GB	262k	how to run ›
185	Qwen3 4B Thinking 2507 Alibaba	46±2†	local	~3.9GB	262k	LM Studio ›
186	Qwen3-VL 235B-A22B Instruct Alibaba	46±6*†	local	~134GB	256k	LM Studio ›
188	Nemotron-Cascade-14B-Thinking NVIDIA	46±4*†	local	~10GB		LM Studio ›
189	Qwen3 14B Alibaba	46±1	local	~10GB	33k	LM Studio ›
191	Qwen3 Coder 30B-A3B Instruct Alibaba	45±2	local	~20GB	160k	LM Studio ›
193	DeepSeek V3 0324 DeepSeek	45±1	local	~108GB	164k	llama.cpp ›
194	Qwen3 30B-A3B Thinking 2507 Alibaba	45±1	local	~20GB	262k	LM Studio ›
195	Nemotron 3 Nano 30B-A3B NVIDIA	45±3*†	local	~20GB	262k	LM Studio ›
196	Granite 4.2 3B IBM	45±2	local	~3.7GB	131k	LM Studio ›
197	Mellum2 12B-A2.5B Instruct JetBrains	45±4*†	local	~8.6GB	131k	LM Studio ›
198	Nemotron Nano 9B v2 NVIDIA	45±4*†	local	~7.6GB	131k	llama.cpp ›
200	Qwen3-VL 4B Thinking Alibaba	45±3*†	local	~4.1GB	256k	llama.cpp ›
202	EXAONE 4.0 32B LG AI Research	45±1	local	~20GB	131k	LM Studio ›
203	Qwen3-VL 30B-A3B Thinking Alibaba	44±2*†	local	~21GB	256k	LM Studio ›
204	Qwen3-VL 30B-A3B Instruct Alibaba	44±3*†	local	~21GB	256k	LM Studio ›
205	Qwen3 32B Alibaba	44±1	local	~20GB	131k	LM Studio ›	+1
206	Command A (03-2025) Cohere	44±1	local	~37GB	256k	llama.cpp ›
207	Falcon-H1R 7B TII	44±2*†	local	~6GB	262k	LM Studio ›
208	Qwen3 235B-A22B (Apr 2025, hybrid) Alibaba	44±1	local	~76GB	131k	llama.cpp ›	+1
209	Qwen3 8B Alibaba	44±1	local	~6.1GB	33k	LM Studio ›	+1
212	Qwen3-Next 80B-A3B Instruct Alibaba	43±2†	local	~27GB	131k	llama.cpp ›
213	Qwen3-VL 8B Instruct Alibaba	43±3*	local	~6.8GB	256k	llama.cpp ›
214	DeepSeek R1 0528 DeepSeek	43±1	local	~110GB	164k	llama.cpp ›
215	Nemotron-Cascade-8B NVIDIA	43±4*†	local	~6.3GB		LM Studio ›
217	Apriel 1.6 15B Thinker ServiceNow	43±4†	local	~10GB		LM Studio ›
218	Spark-X2.5 1.7B iFLYTEK (SparkLLM)	43±2†	local	~2.7GB	1M	llama.cpp ›
219	Qwen2.5 72B Instruct Alibaba	43±1	local	~30GB	33k	llama.cpp ›
220	Qwen2.5 32B Instruct Alibaba	43±1*	local	~21GB	131k	LM Studio ›
222	Qwen3-VL 4B Instruct Alibaba	42±3*†	local	~4.1GB	256k	llama.cpp ›
223	Qwen3 30B-A3B Instruct 2507 Alibaba	42±1	local	~20GB	262k	LM Studio ›
226	Qwen3-VL 8B Thinking Alibaba	42±3*†	local	~6.8GB	256k	llama.cpp ›
227	gpt-oss-20b OpenAI	42±2	local	~14GB	131k	llama.cpp ›
229	Llama 3.1 405B Instruct Meta	42±2	local	~160GB	128k	llama.cpp ›
230	Ministral 3 14B Mistral	42±3	local	~9.1GB	262k	LM Studio ›
231	Qwen3 30B-A3B (Apr 2025, hybrid) Alibaba	42±1	local	~19GB	33k	LM Studio ›
233	Granite 4.1 8B IBM	41±2*†	local	~7GB	131k	LM Studio ›
235	Granite 4.1 3B IBM	41±5*†	local	~3.7GB		Ollama ›
237	Mistral Large 3 (675B-A41B) Mistral	41±8*	local	~108GB	262k	llama.cpp ›
239	Ministral 3 8B Mistral	41±4*	local	~6.5GB	262k	LM Studio ›
240	Mistral Small 3.2 24B Mistral	41±3*	local	~16GB	128k	LM Studio ›
242	LFM2.5 8B-A1B Liquid AI	41±2	local	~6.5GB	128k	LM Studio ›
243	Qwen3 4B Alibaba	41±1	local	~3.7GB	33k	LM Studio ›
244	Qwen2.5 14B Instruct Alibaba	40±1*	local	~10GB	131k	LM Studio ›
245	LFM2.5 1.2B Thinking Liquid AI	40±4*†	local	~2.2GB	33k	LM Studio ›
247	DeepSeek V3 (Dec '24) DeepSeek	40±1	local	~337GB	131k	llama.cpp ›
248	DeepSeek R1 (Jan '25) DeepSeek	40±1	local	~110GB	128k	llama.cpp ›
249	Mistral Small 3.1 24B (2503) Mistral	40±1	local	~17GB	128k	llama.cpp ›
250	Mistral Large 2 (2411) Mistral	40±1	local	~78GB	128k	llama.cpp ›
251	Mistral Large 2 (2407) Mistral	40±1*	local	~45GB	128k	llama.cpp ›
252	QwQ-32B Alibaba	40±1*	local	~21GB	131k	LM Studio ›
253	Qwen3 4B Instruct 2507 Alibaba	40±1	local	~3.9GB	262k	LM Studio ›
257	Llama 3.1 Nemotron Ultra 253B v1 NVIDIA	39±1*	local	~82GB	128k	llama.cpp ›
259	Gemma 4 E4B Google	39±1	local	~6.9GB	131k	LM Studio ›	+3
260	Phi-3.5-mini-instruct Microsoft	39±11*†	local	~3.7GB	128k	LM Studio ›
261	Qwen3-VL 2B Thinking Alibaba	39±3*†	local	~2.7GB	256k	llama.cpp ›
264	Qwen2.5 7B Instruct Alibaba	38±1	local	~6GB	131k	LM Studio ›
265	Granite 4.0 1B IBM	38±3*	local	~2.4GB	128k	LM Studio ›
266	DeepSeek-Coder-V2 Instruct (0724) DeepSeek	38±1*	local	~151GB		llama.cpp ›
268	Qwen3 1.7B Alibaba	38±1	local	~2.6GB	33k	LM Studio ›
269	Qwen3-VL 2B Instruct Alibaba	38±3*†	local	~2.7GB	256k	llama.cpp ›
271	Gemma 3 27B Google	38±1	local	~18GB	131k	LM Studio ›
272	Ministral 3 14B Reasoning 2512 Mistral	38±5*	local	~10GB	262k	llama.cpp ›
274	Llama 4 Maverick (402B-A17B) Meta	37±1	local	~65GB	1M	llama.cpp ›
277	Llama 3.1 70B Instruct Meta	37±1	local	~27GB	128k	llama.cpp ›
278	Nemotron 3 Nano 4B NVIDIA	37±2	local	~3.8GB	262k	LM Studio ›
280	Sky-T1 32B Preview NovaSky (UC Berkeley)	37±1*	local	~21GB		LM Studio ›
281	Granite 4.0 H 1B IBM	37±3*	local	~2.3GB	131k	LM Studio ›
282	Llama 3.3 70B Instruct Meta	36±1	local	~24GB	128k	llama.cpp ›
284	LFM2.5 1.2B Instruct Liquid AI	36±4*†	local	~2.2GB	33k	LM Studio ›
285	Command R7B (12-2024) Cohere	36±1	local	~6.2GB	128k	LM Studio ›
286	Mellum2 12B-A2.5B Thinking JetBrains	36±3†	local	~8.6GB	131k	LM Studio ›
287	Gemma 3 12B Google	36±1*	local	~9.1GB	131k	LM Studio ›
288	Llama 3 70B Instruct Meta	36±1*	local	~24GB	8k	llama.cpp ›
290	Ministral 8B (2410) Mistral	36±1*	local	~6.2GB	128k	LM Studio ›
291	Qwen2.5 3B Instruct Alibaba	36±1*	local	~3.3GB	33k	LM Studio ›
292	Mistral Small Instruct 22B (2409) Mistral AI	36±4*	local	~15GB	32k	LM Studio ›
293	Mistral NeMo Instruct (2407) Mistral	35±1	local	~8.7GB	128k	LM Studio ›
294	Granite 3.1 8B Instruct IBM	35±4*	local	~6.6GB	128k	llama.cpp ›
295	MiniCPM3-4B OpenBMB	35±1*	local	~4GB	32k	LM Studio ›
296	Granite 3.2 8B Instruct IBM	35±4*	local	~6.6GB	131k	llama.cpp ›
297	Llama 3.1 8B Instruct Meta	35±1	local	~6.2GB	128k	LM Studio ›
298	Command R Plus (04-2024) Cohere	34±1*	local	~43GB	128k	llama.cpp ›
299	DeepSeek R1 0528 Qwen3 8B DeepSeek	34±5*	local	~6.3GB	128k	LM Studio ›
300	Qwen2.5 1.5B Instruct Alibaba	34±1*	local	~2.4GB	33k	LM Studio ›
301	Phi-4 14B Microsoft	34±1*	local	~10GB	16k	LM Studio ›
302	Command R Plus (08-2024) Cohere	34±7*†	local	~63GB	128k	LM Studio ›
303	Qwen3 0.6B Alibaba	34±1*	local	~1.8GB	33k	LM Studio ›
304	Hermes 2 Pro Llama 3 8B Nous Research	34±1*	local	~6.6GB	8k	llama.cpp ›
305	Granite 4.0 H 350M IBM	34±5*†	local	~1.7GB	33k	LM Studio ›
307	EXAONE 4.0 1.2B LG AI Research	33±3†	local	~2.2GB	66k	LM Studio ›
308	Falcon3 10B Instruct TII	33±1*	local	~7.6GB	32k	LM Studio ›
309	Llama 4 Scout (109B-A17B) Meta	33±1	local	~36GB	328k	llama.cpp ›
310	Llama 3.2 3B Instruct Meta	33±1	local	~3.4GB	131k	LM Studio ›
312	Gemma 4 E2B Google	32±2	local	~5.1GB	131k	llama.cpp ›	+2
313	Falcon3 7B Instruct TII	32±1*	local	~5.9GB	32k	LM Studio ›
314	Qwen3.5 2B Alibaba	32±2	local	~2.9GB	262k	llama.cpp ›
315	Magistral Small 1.0 (2506) Mistral	31±7*	local	~16GB	128k	LM Studio ›
316	Mixtral 8x7B Instruct v0.1 Mistral	31±1*	local	~29GB	32k	LM Studio ›
317	Mixtral 8x22B Instruct v0.1 Mistral	31±1*	local	~56GB	64k	llama.cpp ›
318	Qwen2 7B Instruct Alibaba	31±1*	local	~6GB	131k	LM Studio ›
319	Bielik 11B v2.3 Instruct SpeakLeash	30±1*	local	~8.6GB		llama.cpp ›
320	Command R (08-2024) Cohere	30±8*†	local	~21GB	128k	LM Studio ›
321	Ministral 3 3B Mistral	30±6*	local	~3.5GB	262k	LM Studio ›
322	ChatGLM3-6B Zhipu AI (THUDM)	30±10*†	local	~5.8GB		llama.cpp ›
323	GLM-4 9B Chat Zhipu AI (Z.ai)	30±1	local	~7.9GB	128k	Ollama ›
324	Gemma 3 4B Google	30±1*	local	~3.8GB	131k	LM Studio ›
325	DBRX Instruct Databricks	30±1*	local	~85GB	32k	Ollama ›
326	Granite 4.0 350M IBM	30±3*	local	~1.7GB	32k	LM Studio ›
327	Olmo 3 7B Instruct Ai2	29±4	local	~5.8GB	66k	LM Studio ›
329	NexusRaven V2 13B Nexusflow	28±1*	local	~9.3GB	16k	Ollama ›
330	Qwen2.5 0.5B Instruct Alibaba	28±1*	local	~1.8GB	33k	LM Studio ›
331	Qwen3.5 0.8B Alibaba	27±5*†	local	~2GB	262k	llama.cpp ›
332	Llama 3 8B Instruct Meta	26±1*	local	~6.6GB	8k	llama.cpp ›
333	Falcon3 3B Instruct TII	26±1*	local	~3.4GB	32k	LM Studio ›
334	Phi-4-mini 3.8B Microsoft	23±1	local	~3.8GB	128k	LM Studio ›
335	Qwen2 1.5B Instruct Alibaba	23±2*	local	~2.4GB	32k	LM Studio ›
336	Llama 3.2 1B Instruct Meta	18±1	local	~2.2GB	131k	LM Studio ›
337	DeepSeek-Coder-V2 Lite Instruct (16B-A2.4B) DeepSeek	18±2*	local	~11GB	128k	LM Studio ›
338	QwQ-32B-Preview Alibaba	17±2*	local	~21GB	33k	LM Studio ›
339	Falcon3 1B Instruct TII	17±2*	local	~2.4GB	8k	LM Studio ›
340	Gemma 3 1B Google	13±2	local	~2.2GB	32k	LM Studio ›
—	Devstral 2 (2512) Mistral	not yet calibrated	local	~75GB	256k	LM Studio ›
—	Devstral Small 2 24B Mistral	not yet calibrated	local	~17GB	262k	llama.cpp ›
—	GLM-4.7-Flash Z.ai	not scored yet	local	~19GB	200k	LM Studio ›
—	Granite 4.0 H Micro 3B IBM	not yet calibrated	local	~3.4GB	131k	LM Studio ›
—	Granite 4.0 H Small (32B-A9B) IBM	not yet calibrated	local	~21GB	131k	LM Studio ›
—	Granite 4.0 H Tiny (7B-A1B) IBM	not yet calibrated	local	~5.6GB	131k	LM Studio ›
—	Granite 4.0 Micro 3B IBM	not yet calibrated	local	~3.5GB	131k	LM Studio ›
—	K2 Horizon 7B Institute of Foundation Models (MBZUAI)	not scored yet	local	~7.7GB	524k	llama.cpp ›
—	Laguna M.1 Poolside	not yet calibrated	local	~135GB	262k	LM Studio ›
—	Laguna XS 2.1 (33B-A3B) Poolside	not yet calibrated	local	~21GB	262k	LM Studio ›
—	Laguna XS.2 Poolside	not yet calibrated	local	~21GB	262k	LM Studio ›
—	Ling 3.0 Flash (124B-A5B) inclusionAI (Ant Group)	not yet calibrated	local	~75GB	131k	LM Studio ›
—	Llama 3.3 Nemotron Super 49B v1 NVIDIA	not scored yet	local	~22GB		llama.cpp ›
—	MiniMax-M1 (80k) MiniMax	not yet calibrated	local	~959GB	1M	how to run ›
—	Nemotron Cascade 2 30B-A3B NVIDIA	not scored yet	local	~20GB	256k	LM Studio ›
—	Qwen3-Coder-Next (80B-A3B) Alibaba	not scored yet	local	~27GB	262k	llama.cpp ›
—	Step 3.7 Flash (198B-A11B) StepFun	not scored yet	local	~104GB	256k	llama.cpp ›
—	Trinity-Large-Thinking Arcee AI	not scored yet	local	~141GB	512k	llama.cpp ›

+ 129 more don't fit your machine or budget · show them

+ 33 models with fewer than 3 results left off, since one or two tests say little about overall ability · show them

362 of 398 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