Tier list
Local AI model tier list
Every model we track, ranked by quality alone. Tiers don’t depend on your hardware: they tell you how capable a model is. To see which ones your machine can run, use the Model Finder.
How the 621 LLMs stack up
Top ratedKimi K3.1 Pro (2.8T)99 Best small modelGemma 4 12B83 Newest top-tierLongCat 2.5 (1.6T-A48B)98 Best fully openDeepSeek V4.1 Flash 1017 (552B-A16B)98
Who makes the best ones · models rated A or A+
- Alibaba32
- DeepSeek20
- Zhipu AI14
- Mistral AI13
- NVIDIA11
- Moonshot AI10
Step 1
Tell us your hardware first
Then you will see how many of these 621 LLMs run on it, and can filter the list by how well each one runs.
A+
Excellent · scores 85–100
Frontier class: rivals the best hosted models at reasoning, code and long tasks.
109 models
- Kimi K3.1 Pro (2.8T) 99
- DeepSeek V4.1 Flash 1017 (552B-A16B) 98
- GLM-5.3 (753B-A40B) 98
- Qwen3.8 Max (2.4T-A95B) 98
- Kimi K3 (2.8T) 98
- LongCat 2.5 (1.6T-A48B) 98
- MiMo V2.6 Pro (1T-A42B) 97
- Ling 3.1 Flash 1019 (560B-A25B) 97
- MiniMax M3.1 Flash (428B-A23B) 97
- Nex N2.6 (321B-A18B) 97
- Step 5 Preview (600B-A27B) 97
- GLM-5.3 Flash (320B-A18B) 97
- MiMo V2.6 Flash (310B-A15B) 97
- DeepSeek V4.1 Flash (552B-A16B) 97
- Mistral Large 4 (1T-A52B) 97
- Qwen3.8 Flash Next (125B-A6B) 97
- GLM-5.2 (753B-A40B) 96
- Hunyuan Hy4 Preview (780B) 96
- Muse Spark Open (816B-A42B) 96
- DeepSeek V4 Pro (1.6T-A49B) 96
- Kimi K2.6 (1T) 96
- Nanbeige4.3 33B 96
- Ling 3.1 Flash (560B-A25B) 96
- MiMo V2.5 Pro (1T-A42B) 96
- Qwen3.8 27B 96
- Intern-S2 (397B-A17B) 95
- GLM-5.1 (744B-A40B) 95
- Nex N2 Pro (397B-A17B) 95
- DeepSeek V4 Pro (April 2026) 95
- Naive N0.5 Flash (309B-A15.5B) 95
- Qwen3.5 397B-A17B 95
- MiniMax M3 (427B) 95
- GLM-5 (744B-A40B) 95
- K2-Horizon (375B-A23B) 95
- Hunyuan Hy3 (295B-A21B) 94
- Beam 1.0 (501B-A23B) 94
- Kimi K2.5 (1T) 94
- Kimi K2.7 Code (1T) 94
- DeepSeek V4 Flash (284B-A13B) 94
- MiMo V2.5 (310B-A15B) 94
- Ornith 1.5 397B-A17B 94
- Qwen3.6 27B 94
- Nemotron 3.5 Super VL 120B-A12B 94
- DeepSeek V3.2 (671B) 94
- Inkling Small (276B-A12B) 94
- GLM-4.7 (355B-A32B) 94
- Inkling (975B-A41B) 94
- MiniMax M2.7 (230B-A10B) 94
- Qwen3.6 35B-A3B 94
- Step 3.7 Flash (198B-A11B) 94
- MiniCPM-V 4.7 35B-A3B 93
- Nemotron 3 Ultra 550B-A55B 93
- DeepSeek R1 0528 (671B) 93
- Kimi K2 Thinking (1T) 93
- Qwen3.5 27B 93
- Qwen3.5 35B-A3B 93
- Ornith 1.5 122B-A10B 93
- DeepSeek V3.2 Exp 93
- Gemma 4 26B-A4B 93
- Ling 2.6 1T 93
- MiniMax M2.5 (230B-A10B) 93
- Nemotron 3.5 Bolt 74B-A9B 93
- DeepSeek R1 (671B) 92
- DeepSeek V4 Flash (April 2026) 92
- Muse Glimmer 30B 92
- Nex N2 Mini (35B-A3B) 92
- Qwen3.5 122B-A10B 92
- Step 3.5 Flash (196B-A11B) 92
- DeepSeek V3.1 Terminus 92
- DeepSeek V3.1 (671B) 91
- Gemma 4 31B 91
- KAT-Coder V2.5 Dev (35B-A3B) 91
- Kimi K2 Instruct 0905 (1T) 91
- Kolibri 1 (78B-A3.5B) 91
- MiniMax M2.1 (230B-A10B) 91
- Qwen3 235B-A22B Thinking 2507 91
- Qwen3 Coder 480B-A35B 91
- Sage-A1.5 35B-A3B 91
- MiMo V2 Flash (309B-A15B) 91
- Flint (40B-A2.5B) 90
- Ling 1T 90
- Mistral Large 3 (675B) 90
- Ornith 1.5 35B-A3B 90
- Qwen3 235B-A22B Instruct 2507 90
- Solar Open 2 250B 90
- TMax 27B 90
- Apodex 1.1 Mini (35B-A3B) 89
- Kimi K2 Instruct (July 2025) 89
- MiniMax M2 (230B-A10B) 89
- Ornith 1.0 35B-A3B 89
- Qwen3-VL 235B-A22B Instruct 89
- Command A+ (218B-A25B) 88
- Fara 1.5 27B 88
- GLM-4.7 Flash (30B-A3B) 88
- K-EXAONE 236B-A23B 88
- LongCat Flash Chat (560B) 88
- Nemotron 3 Super 120B-A12B 88
- Qwen3 235B-A22B 88
- Raon-1 mini Preview (30B-A4B) 88
- Agents-A1 35B-A3B 87
- Qwen AgentWorld 35B-A3B 87
- Devstral 2 (123B) 86
- InternVL3.5 241B-A28B 86
- Laguna S 2.1 (118B-A8B) 86
- Llama 3.1 Nemotron Ultra 253B v1 86
- Qwen3 Coder Next (80B-A3B) 86
- Qwen3-Next 80B-A3B Thinking 86
- UI-Mate 27B 86
- Hunyuan Hy-Memory 30B-A3B 85
A
Great · scores 70–84
Great everyday models: strong at coding, writing and tool use.
117 models
- Baichuan M3 235B-A22B 84
- GLM-4.6 (355B-A32B) 84
- gpt-oss 120B 84
- Qwen3 Next 80B-A3B Instruct 84
- DeepSeek V3 0324 (671B) 83
- Gemma 4 12B 83
- GLM-4.6V 83
- MiniMax-M1 80K 83
- Command A Reasoning (111B) 82
- EXAONE 4.5 33B 82
- GLM-4.5 (355B-A32B) 82
- Laguna XS 2.1 (33B-A3B) 82
- Ling 3.0 Flash (127B-A5B) 82
- MiniMax-M1 40K 82
- Mistral Small 4 (119B-A6B) 81
- Mistral Medium 3.5 (128B) 80
- Trinity Large Thinking (400B-A13B) 80
- Command A (111B) 80
- ERNIE 4.5 300B-A47B 80
- GLM-4.5V 80
- Hermes 4 405B 80
- K2 Horizon 7B 80
- Kimi Dev 72B 80
- Ling 2.6 Flash (104B-A7.4B) 80
- Llama 3.3 Nemotron Super 49B v1.5 80
- Mistral Large 2.1 (2411) 80
- Nemotron Cascade 2 30B-A3B 80
- North Mini Code 1.0 (30B-A3B) 80
- Seed-OSS 36B Instruct 80
- Solar Open 100B 80
- Hermes 4.3 36B 79
- DiffusionGemma 26B-A4B IT 78
- GLM-4.5 Air (106B-A12B) 78
- MiMo V2.6 Distill Qwen 9B 78
- Mistral Large 2 (123B) 78
- NorthCUA Pro 78
- Pixtral Large 124B (2411) 78
- Qwen3 30B-A3B Thinking 2507 78
- Qwen3 Coder 30B-A3B 78
- Qwen3-VL 32B Instruct 78
- QwQ 32B 78
- Ring Flash 2.0 (100B-A6B) 78
- SenseNova Loop 9B Preview 78
- Nemotron 3 Nano 30B-A3B 77
- Ornith 1.5 9B 77
- Qwen3 30B-A3B Instruct 2507 77
- Devstral Small 2 (24B) 76
- dots.llm1 Instruct 76
- Granite 4.2 30B 76
- InternVL3.5 38B 76
- Ling Flash 2.0 (100B-A6B) 76
- MiniMax-Text-01 76
- Olmo 3.1 32B Think 76
- OpenReasoning Nemotron 32B 76
- Qwen3 32B 76
- Qwen3-VL 30B-A3B Thinking 76
- Qwen3.5 9B 76
- Skywork OR1 32B 76
- Tülu 3 405B 76
- DeepSeek V3 (671B) 75
- Ornith 1.0 9B 75
- Sarvam 105B 75
- DeepSeek Coder V2 (236B) 74
- EXAONE Deep 32B 74
- GLM-4 32B 0414 74
- GLM-Z1 32B (0414) 74
- Hermes 4 70B 74
- Hunyuan A13B (80B-A13B) 74
- InternVL3 78B 74
- Kimi Linear 48B-A3B 74
- Leanstral 119B-A6B 74
- LFM2.5 24B-A2B 74
- Llama 3.3 70B Instruct 74
- Qwen3 30B-A3B 74
- Qwen3-VL 30B-A3B Instruct 74
- TMax 9B 74
- Qwen2.5 72B Instruct 73
- Qwen2.5-VL 72B Instruct 73
- Apriel 1.5 15B Thinker 72
- DeepSeek Coder V2 Instruct (0614) 72
- DeepSeek R1 Distill Llama 70B 72
- DeepSeek R1 Distill Qwen 32B 72
- DeepSeek V2.5 1210 72
- EXAONE 4.0 32B 72
- Gemma 3 27B IT 72
- Granite 4.1 30B 72
- Llama 3.1 405B Instruct 72
- Llama 4 Maverick (17B x 128E) 72
- Llama 4 Scout 17B-16E 72
- Magistral Small 1.2 (24B) 72
- Nemotron 3.5 Lightning 30B-A3B 72
- Olmo 3.1 32B Instruct 72
- Phi-4 Reasoning Plus (14B) 72
- Tülu 3 70B 72
- Baichuan M2 32B 70
- Command R+ (08-2024) 104B 70
- DeepSeek V2.5 70
- Devstral Small 1.1 (24B) 70
- ERNIE 4.5 21B-A3B Thinking 70
- gpt-oss 20B 70
- Groot 65B A6B 261023 70
- InternVL3 38B 70
- Jamba 1.5 Large 70
- Llama 3.1 70B Instruct 70
- Magistral Small 2507 70
- Maple Preview (20B-A1B) 70
- MedGemma 27B Text IT 70
- Mellum2.1 Thinking (12B-A2.5B) 70
- Nemotron 3.5 Bolt E4B 70
- OpenHands LM 32B 70
- Phi-4 Reasoning (14B) 70
- Phi-4 Reasoning Vision 15B 70
- Qwen2.5 32B Instruct 70
- Qwen3-Omni 30B-A3B Instruct 70
- Sarvam 30B 70
- UI-Mate 9B 70
- UI-Venus 2 9B 70
B
Good · scores 55–69
Good, dependable models for chat, drafting and summaries.
97 models
- Gemma 4 E4B 68
- Granite 4.2 8B 68
- Hermes 3 (Llama 3.1 70B) 68
- Ling 3.0 flash Sante 68
- Magistral Small 2506 68
- Mistral Small 3.2 24B 68
- Nanbeige4.2 3B 68
- Nemotron Labs Dory 1.0 4B 68
- QVQ 72B Preview 68
- Qwen2.5-VL 32B Instruct 68
- Qwen3 14B 68
- Terminal RSI 27B 68
- Hemmingway 1.5 67
- Qwen2.5 Coder 32B Instruct 67
- AceReason Nemotron 14B 66
- Apertus v1.5 70B 66
- Apriel Nemotron 15B Thinker 66
- Devstral Small 2505 66
- Hermes 4 14B 66
- Llama 3.1 Nemotron 70B Instruct 66
- Llama 3.2 90B Vision Instruct 66
- Ministral 3 14B Reasoning 66
- Mistral Small 3.1 24B 66
- OpenReasoning Nemotron 14B 66
- Qwen2.5 Coder 14B Instruct 66
- QwQ 32B Preview 66
- Reka Flash 3.1 (21B) 66
- SenseNova 6.8 Code2Video 66
- Spark-X2.6 4B 66
- Trinity Mini (26B-A3B) 66
- Command R+ 104B (04-2024) 64
- DeepCoder 14B Preview 64
- ERNIE 4.5 21B-A3B 64
- Falcon-H1 34B Instruct 64
- InternVL3 14B 64
- Ministral 3 14B 64
- Mistral Small 3 24B (2501) 64
- Phi-4 (14B) 64
- Qwen2 72B Instruct 64
- Qwen2-VL 72B Instruct 64
- Axion 1.0 35B 62
- Codestral 22B 62
- Command R (08-2024) 32B 62
- DeepSeek R1 0528 Qwen3 8B 62
- DeepSeek R1 Distill Qwen 14B 62
- DeepSeek V2 Chat 62
- Dolphin Mistral 24B Venice Edition 62
- Gemma 2 27B IT 62
- Gemma 4 E2B 62
- Granite 4.0 H Small (32B-A9B) 62
- InternVL3.5 8B 62
- JanusCoder 14B 62
- Qwen2.5 14B Instruct 62
- Qwen3-VL 8B Instruct 62
- Qwen3.5 4B 62
- Voxtral Small 24B 62
- TMax 4B 61
- Agents-A1 4B 60
- Apertus 70B Instruct 60
- Aya Expanse 32B 60
- Aya Vision 32B 60
- Dolphin 3.0 R1 Mistral 24B 60
- Gemma 3 12B IT 60
- GLM-4.1V 9B Thinking 60
- Granite 4.1 8B 60
- Kimi-VL A3B Thinking 2506 60
- Ling 3.0 Tiny (8B-A1.3B) 60
- Mage-VL-Cowork 4.7B 60
- MiniCPM-V 4.5 8B 60
- MiniCPM5 2B 60
- Nemotron Nano 9B v2 60
- OLMo 2 32B Instruct 60
- Qwen2.5 14B Instruct 1M 60
- Qwen3 8B 60
- Solar Pro Preview 22B 60
- NorthCUA Mini 59
- Fara 1.5 4B 58
- GLM-Z1 9B (0414) 58
- Llama 3 70B Instruct 58
- LLM4Decompile 22B V2 58
- MiMo-VL 7B RL 58
- Qwen2.5-Math 72B Instruct 58
- Seed-Coder 8B Reasoning 58
- Groot 19B A3B 261023 57
- EXAONE Deep 7.8B 56
- Falcon-H1R 7B 56
- InternVL3 8B 56
- LFM2 24B-A2B 56
- LFM3 1.7B 56
- Mage-VL 4.7B 56
- MiMo 7B RL 56
- Ministral 3 8B 56
- Phi-3.5 MoE (42B-A6.6B) 56
- Qwen3 4B Thinking 2507 56
- Seed-Coder 8B Instruct 56
- Mixtral 8x22B Instruct 55
- Qwen2.5 Coder 7B Instruct 55
C
Fair · scores 40–54
Fair: fine for simple tasks, classification and quick answers.
132 models
- Command R 35B (v01) 54
- Llama 3.1 Nemotron Nano 8B v1 54
- MiniCPM4 8B 54
- Mistral Small 22B (2409) 54
- Olmo 3 7B Think 54
- OpenReasoning Nemotron 7B 54
- Qwen3 4B Instruct 2507 54
- Qwen3-VL 4B Instruct 54
- DeepSeek Coder V2 Lite (16B) 52
- Falcon 3 10B Instruct 52
- Foundation-Sec 8B Reasoning 52
- GLM-4 9B 0414 52
- Granite 4.2 3B 52
- Hunyuan 7B Instruct 52
- LFM2.5 8B-A1B 52
- Ling Mini 2.0 (16B-A1.4B) 52
- MiniCPM-o 2.6 8B 52
- Mistral Nemo 12B Instruct 52
- NorthCUA Nano 52
- Nous Hermes 2 Mixtral 8x7B DPO 52
- Qwen1.5 110B Chat 52
- Qwen2.5 7B Instruct 52
- Qwen2.5-VL 7B Instruct 52
- Aya 23 35B 50
- DeepSeek R1 Distill Llama 8B 50
- Falcon-H1 7B Instruct 50
- Fara 7B 50
- Foundation-Sec 1.1 8B Instruct 50
- Gemma 2 9B IT 50
- GLiDE 9B 50
- Grok-1 (314B) 50
- InternLM2.5 20B Chat 50
- InternLM3 8B Instruct 50
- InternVL2.5 8B 50
- Jamba 1.5 Mini 50
- Kimi-VL A3B Instruct 50
- LFM3-VL 2B 50
- Mixtral 8x7B Instruct 50
- Nemotron 3 Nano 4B 50
- Olmo 3 7B Instruct 50
- Pixtral 12B (2409) 50
- Qianfan-OCR 4B 50
- Qwen2 57B-A14B Instruct 50
- Qwen2.5 7B Instruct 1M 50
- Qwen3 4B 50
- Snowflake Arctic Instruct 50
- Yi 1.5 34B Chat 50
- Chandra OCR 2 48
- GLM-4 9B Chat (2024) 48
- GLM-4 9B Chat 1M 48
- Hermes 3 (Llama 3.1 8B) 48
- LightOnOCR-3 4B 48
- Llama 3.1 8B Instruct 48
- MiniCPM-V 2.6 8B 48
- Apertus v1.5 8B 46
- CodeGeeX4 All 9B 46
- DeepHermes 3 Llama 3 8B Preview 46
- DeepSeek R1 Distill Qwen 7B 46
- Dolphin 3.0 Llama 3.1 8B 46
- EXAONE 3.5 7.8B 46
- Falcon 3 7B Instruct 46
- Gemma 3n E4B 46
- Granite 3.2 8B Instruct 46
- Granite 3.3 8B 46
- Granite 4.1 3B 46
- Ministral 8B (2410) 46
- Moonlight 16B-A3B Instruct 46
- Phi-3 Medium (14B) 46
- Phi-3 Medium 128K Instruct 46
- Phi-4 Mini Reasoning (3.8B) 46
- Qwen2.5-Omni 7B 46
- Tülu 3 8B 46
- Gemma 3 4B IT 45
- WeVisOne-VL 4B 45
- Aya Vision 8B 44
- Code Llama 70B Instruct 44
- Command R7B 44
- DeepSeek Coder 33B Instruct 44
- Dolphin 2.9.4 Llama 3.1 8B 44
- Granite 3.1 8B Instruct 44
- Granite 4.0 H Tiny (7B-A1B) 44
- Hermes 2 Pro Llama 3 8B 44
- InternVL2 8B 44
- Jamba Reasoning 3B 44
- LFM2.5 2.6B 44
- LFM2.5-VL 3B 44
- Llama 3.1 Nemotron Nano 4B v1.1 44
- LLaVA 1.6 34B 44
- MedGemma 4B IT 44
- Ministral 3 3B 44
- North Micro Vision 44
- OLMo 2 13B Instruct 44
- OpenChat 3.6 8B 44
- Phi-3 Small 8K Instruct 44
- Phi-4 Mini (3.8B) 44
- Phi-4 Multimodal Instruct 44
- Qwen1.5 32B Chat 44
- Qwen2 7B Instruct 44
- Qwen2-VL 7B Instruct 44
- Unlimited OCR 2 (3B-A0.6B) 44
- Command R7B Arabic 42
- Hunyuan 4B Instruct 42
- InternLM2 Chat 20B 42
- Jamba2 3B 42
- Llama 3 8B Instruct 42
- Llama 3.2 11B Vision Instruct 42
- MiniCPM-OCR 1.0 1.3B 42
- Mistral-NeMo-Minitron 8B Instruct 42
- PaddleOCR-VL 1.6 0.9B 42
- SmolLM3 3B 42
- AFM 4.5B 40
- Apertus 8B Instruct 40
- Aya Expanse 8B 40
- DeepSeek LLM 67B Chat 40
- DeepSeek Prover V2 7B 40
- DeepSeek V2 Lite Chat 40
- Granite 3.0 8B Instruct 40
- Granite 4.0 Micro (3B) 40
- Granite Code 34B Instruct 40
- InternLM2 Math Plus 20B 40
- InternLM2.5 7B Chat 40
- MiniCPM-V 4.7 1B 40
- Nous Hermes 2 SOLAR 10.7B 40
- Phi-3.5 Mini Instruct 40
- Qwen3.5 2B 40
- SOLAR 10.7B Instruct 40
- StarCoder2 15B Instruct 40
- TeleOCR 40
- TMax 2B 40
- Unlimited OCR (3B-A0.6B) 40
- Yi 34B Chat 40
- Yi Coder 9B Chat 40
D
Basic · scores under 40
Basic: tiny or older models for experiments and small devices.
166 models
- Code Llama 34B Instruct 38
- CogVLM2 Llama 3 Chat 19B 38
- Dolphin 2.9 Llama 3 8B 38
- EXAONE 3.0 7.8B Instruct 38
- Falcon3 Mamba 7B Instruct 38
- Granite 4.0 H Micro 38
- LFM2 8B-A1B 38
- LLM4Decompile 9B V2 38
- MiniCPM3 4B 38
- Mistral 7B Instruct v0.3 38
- Phi-3.5 Vision Instruct 38
- Zamba2 7B Instruct 38
- S5 nano 37
- Codestral Mamba 7B 36
- Falcon-H1 3B Instruct 36
- Gemma 3n E2B 36
- Granite Code 20B Instruct 36
- Hermes 2 Pro Mistral 7B 36
- Llama 2 70B Chat 36
- MiniCPM5 1B 36
- OLMo 2 7B Instruct 36
- Phi-3 Mini 128K Instruct 36
- Phi-3 Mini 4K Instruct 36
- Qwen2.5 Coder 3B Instruct 36
- Qwen2.5-Math 7B Instruct 36
- Qwen3-VL 2B Instruct 36
- Starling LM 7B Beta 36
- Voxtral Mini 3B 36
- Yi 1.5 9B Chat 36
- Llama 3.2 3B Instruct 35
- OpenChat 3.5 (0106) 35
- Aya 23 8B 34
- CodeGemma 7B 34
- DeepSeek Coder 6.7B Instruct 34
- Falcon 180B Chat 34
- LFM2 2.6B 34
- Mathstral 7B 34
- Mistral 7B Instruct v0.2 34
- Nous Hermes 2 Mistral 7B DPO 34
- OpenChat 3.5 (1210) 34
- Phi-3 Vision 128K Instruct 34
- Qwen1.5 14B Chat 34
- Qwen2.5 3B Instruct 34
- Qwen2.5-VL 3B Instruct 34
- RecurrentGemma 9B IT 34
- Granite Code 8B Instruct (128K) 32
- Hermes 3 Llama 3.2 3B 32
- InternLM2 Math Plus 7B 32
- LFM2.5 1.2B Instruct 32
- LLM4Decompile 6.7B V2 32
- OpenHermes 2.5 Mistral 7B 32
- StarCoder2 15B (base) 32
- Starling LM 7B Alpha 32
- Code Llama 13B Instruct 30
- EXAONE 3.5 2.4B 30
- EXAONE 4.0 1.2B 30
- Falcon 3 3B Instruct 30
- Falcon Mamba 7B Instruct 30
- Granite 3.2 2B Instruct 30
- Granite 3.3 2B 30
- Mistral 7B Instruct v0.1 30
- Qwen2.5-Omni 3B 30
- Qwen3 1.7B 30
- SQLCoder 7B v2 30
- Gemma 1.1 7B IT 28
- Gemma 2 2B 28
- Granite 3.1 2B Instruct 28
- Granite 4.0 H 1B 28
- LLaVA 1.6 Mistral 7B 28
- Neural Chat 7B v3.3 28
- Orca 2 13B 28
- StableLM 2 12B Chat 28
- Yi 1.5 6B Chat 28
- WeVisOne VL 2B 27
- Baichuan 2 13B Chat 26
- DeepScaleR 1.5B Preview 26
- DeepSeek MoE 16B Chat 26
- Falcon-H1 1.5B Instruct 26
- Gemma 7B IT 26
- Granite 3.0 2B Instruct 26
- Granite Vision 3.2 2B 26
- LFM2 1.2B 26
- Llama 2 13B Chat 26
- OLMoE 1B-7B Instruct 26
- Qwen1.5 7B Chat 26
- Qwen2.5 Coder 1.5B Instruct 26
- Qwen3.5 0.8B 26
- StarCoder 15B 26
- WizardMath 7B v1.1 26
- Zephyr 7B Beta 26
- Code Llama 7B Instruct 24
- LFM2-VL 1.6B 24
- Qwen1.5 MoE A2.7B Chat 24
- Qwen2-VL 2B Instruct 24
- Qwen2.5 1.5B Instruct 24
- SmolVLM2 2.2B Instruct 24
- StarCoder2 7B 24
- DeepSeek R1 Distill Qwen 1.5B 22
- Hunyuan 1.8B Instruct 22
- Llama 3.2 1B Instruct 22
- LLaVA 1.5 13B 22
- Nemotron Mini 4B Instruct 22
- Orca 2 7B 22
- SmolVLM 2.2B Instruct 22
- WizardLM 13B v1.2 22
- Zamba2 2.7B Instruct 22
- d1 2B 21
- Baichuan 2 7B Chat 20
- Gemma 3 1B 20
- Granite 3.1 3B-A800M Instruct 20
- Granite Code 3B Instruct (128K) 20
- Llama 2 7B Chat 20
- Qwen2.5-Math 1.5B Instruct 20
- SmolLM2 1.7B Instruct 20
- Stable Code Instruct 3B 20
- Vicuna 13B v1.5 20
- Yi 6B Chat 20
- DeepSeek LLM 7B Chat 18
- Granite 3.0 3B-A800M Instruct 18
- InternLM2.5 1.8B Chat 18
- LFM2 700M 18
- StarCoder2 3B 18
- GLiDE 2B 17
- Falcon 3 1B Instruct 16
- Falcon 40B Instruct 16
- LLaVA 1.5 7B 16
- Phi-2 2.7B 16
- Qwen2 1.5B Instruct 16
- Spark X2.6 1.7B 16
- StableLM Zephyr 3B 16
- Yi Coder 1.5B Chat 16
- surya ocr 2.5 15
- CodeGemma 2B 14
- Falcon-H1 0.5B Instruct 14
- LFM2.5 350M 14
- OLMo 2 1B Instruct 14
- OLMo 7B Instruct 14
- Qwen3 0.6B 14
- SmolLM 1.7B Instruct 14
- Vicuna 7B v1.5 14
- DeepSeek Coder 1.3B Instruct 12
- Gemma 2B IT 12
- Granite 4.0 H 350M 12
- InternLM 7B 12
- LFM2 350M 12
- LFM2-VL 450M 12
- LLM4Decompile 1.3B V2 12
- GLiDE 0.8B 11
- LFM2.5 230M 10
- Qwen2.5 0.5B Instruct 10
- SmolVLM 500M Instruct 10
- StableLM 2 1.6B Chat 10
- StableLM 2 Zephyr 1.6B 10
- Hunyuan 0.5B Instruct 8
- Qwen2.5 Coder 0.5B Instruct 8
- SmolLM2 360M Instruct 8
- ERNIE 4.5 0.3B 6
- Falcon 7B Instruct 6
- Gemma 3 270M 6
- Qwen2 0.5B Instruct 6
- SmolVLM 256M Instruct 6
- TinyLlama 1.1B Chat 6
- SmolLM 360M Instruct 5
- SmolLM2 135M Instruct 4
- Falcon-H1 Tiny 90M Instruct 3
- SmolLM 135M Instruct 2
Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.