Best Local LLMs for Coding in 2026
Best Local LLM for Coding by VRAM
8 GB
RTX 5060
AlternativeNo qualifying model yet12 GB
RTX 5070
16 GB
RTX 5080
24–48+ GB
RTX 4090
Highest llmrun Coding Score among official releases that fit each memory size at a Q3-or-better quant; speed is decode tok/s on that size's reference GPU. Full coding ranking
These are the open-weight code models you can run on your own hardware — ranked by their llmrun Score for coding, with the best pick for every VRAM size first. Local coding models keep your codebase private, work offline, and cost nothing per token. For most developers a 7B–32B model at Q4_K_M is the sweet spot: small enough to fit a single consumer GPU, capable enough for autocomplete, refactors, and agentic coding. Pick a model below to see exactly which GPU or Mac runs it and how fast.
135 Coding Models You Can Run Locally
Top 11 by llmrun Coding Score, official releases that fit 48 GB or less
- 1
Qwen3.8 27BAlibaba · 27.8B
Alibaba's Qwen3.8 27B is a dense, vision-capable model with around 28 billion parameters, continuing the Qwen line into its 3.8 release. It can process images alongside text prompts, supporting tasks like visual question answering and image-grounded chat in addition to…
- 8
- 12
- 16
- 24
- 32
- 48
120
Q3_K_L · 15.4 GB
42 tok/s
- 2
Gemma 4 31B ITGoogle · 31.3B
Gemma 4 31B IT is Google's 31-billion-parameter instruction-tuned model in the Gemma 4 lineup, built to handle both text and image input in a single pass. As a vision-capable model, it can describe, compare, or reason about images alongside written…
- 8
- 12
- 16
- 24
- 32
- 48
108
Q4_K_M · 21.7 GB
32 tok/s
- 3
Qwen3.6 35B A3BAlibaba · 36.0B
Qwen3.6 35B A3B pairs a mixture-of-experts design with roughly 35 billion total parameters, only about 3 billion of which are active per token. That small active footprint keeps generation speed close to a 3B-class dense model, even though the full…
- 8
- 12
- 16
- 24
- 32
- 48
105
IQ3_S · 15.7 GB
181 tok/s
- 4
Gemma 4 26B A4B ITGoogle · 25.8B
Gemma 4 26B A4B IT applies a mixture-of-experts design to Google's Gemma 4 family, with roughly 26 billion total parameters but only about 4 billion active for any given token. That active-parameter count is what determines inference speed, so despite…
- 8
- 12
- 16
- 24
- 32
- 48
104
IQ3_S · 12 GB
162 tok/s
- 5
Qwen3.6 27BAlibaba · 27.8B
Released a few months before its Qwen3.8 sibling, Qwen3.6 27B is Alibaba's dense, roughly 28-billion-parameter model from the Qwen 3.6 generation. Vision input is supported alongside text, so the model can handle image-based prompts as well as standard chat, coding…
- 8
- 12
- 16
- 24
- 32
- 48
102
Q3_K_L · 15.4 GB
42 tok/s
- 6
GPT OSS 20BOpenAI · 20.9B
GPT-OSS 20B is one of OpenAI's first open-source model releases, marking a historic shift in the company's approach to open weights. At 21.5 billion parameters it delivers strong general-purpose chat and reasoning capabilities informed by the research behind the GPT…
- 8
- 12
- 16
- 24
- 32
- 48
101
IQ4_XS · 11.8 GB
150 tok/s
- 7
Qwen3.5 27BAlibaba · 27.8B
Qwen3.5 27B is a 27.8-billion-parameter dense model from Alibaba's Qwen team, built for both text and image input using the same hybrid linear/full-attention architecture as the rest of the Qwen3.5 family. As a vision-capable model it can describe and reason…
- 8
- 12
- 16
- 24
- 32
- 48
98
Q3_K_L · 15.4 GB
42 tok/s
- 8
Qwen3 32BAlibaba · 32.8B
Qwen3 32B is the flagship dense model in Alibaba Cloud's Qwen 3 series, with 32 billion parameters. It is instruction-tuned for chat and delivers strong performance across reasoning, coding, mathematics, and multilingual tasks.
- 8
- 12
- 16
- 24
- 32
- 48
93
Q4_K_M · 22.6 GB
32 tok/s
- 9
Llama 3.3 70B InstructMeta · 70.6B
Meta Llama 3.3 70B Instruct is a 70-billion parameter large language model from Meta, released as part of the Llama 3.3 generation. It is an instruction-tuned model optimized for dialogue and chat use cases, offering strong performance across reasoning, coding…
- 8
- 12
- 16
- 24
- 32
- 48
76
IQ4_XS · 46.4 GB
15 tok/s
- 10
Gemma 3 27B ITGoogle · 27.4B
Google Gemma 3 27B IT is a 27.4-billion parameter multimodal instruction-tuned model from Google's Gemma 3 family. It supports both text and image inputs, making it one of the most capable openly available vision-language models for local inference.
- 8
- 12
- 16
- 24
- 32
- 48
70
Q4_K_M · 21.7 GB
36 tok/s
- 11
Llama 3.1 8B InstructMeta · 8.0B
Meta Llama 3.1 8B Instruct is an 8-billion parameter instruction-tuned language model from Meta. Part of the Llama 3.1 release, it supports a 128K token context window and is fine-tuned for conversational use, tool calling, and general assistant tasks.
- 8
- 12
- 16
- 24
- 32
- 48
56
Q4_K_M · 7.9 GB
55 tok/s
All 135 models tagged for coding
- Qwen2.5 Coder 14B Instruct12 GB+12.4 GB @Q4
Alibaba · 14.8B
Qwen2.5 Coder 14B Instruct is a 14.8B-parameter open language model from Alibaba in the Qwen 2.5 family.
- Qwen2.5 Coder 7B Instruct8 GB+5.8 GB @Q4
Alibaba · 7.6B
Qwen2.5 Coder 7B Instruct is a 7.6-billion parameter code-specialized instruction-tuned model from Alibaba Cloud.
- Qwen2.5 Coder 32B Instruct24 GB+24.3 GB @Q4
Alibaba · 32.8B
Qwen2.5 Coder 32B Instruct is a 32.8-billion parameter code-specialized model from Alibaba Cloud, instruction-tuned for programming assistance and code generation.
- DeepSeek Coder v2 Lite Instruct8 GB+10.2 GB @Q4
DeepSeek · 15.7B
DeepSeek Coder V2 Lite Instruct is a code-focused mixture-of-experts model with 15.7 billion total parameters, trained to handle both programming tasks and general conversation.
- Phi 3.5 Vision Instruct12 GB+9.2 GB @Q4
Microsoft · 4.1B
Phi 3.5 Vision Instruct is Microsoft's 4.1-billion-parameter multimodal model, pairing an image encoder and connector with the Phi-3 Mini language model to handle text and…
- Deepseek Coder 7B Instruct V1.58 GB+6.5 GB @Q4
DeepSeek · 6.9B
Deepseek Coder 7B Instruct V1.5 is a 6.9-billion-parameter code-focused language model from DeepSeek, fine-tuned from the DeepSeek-LLM 7B base for programming assistance, code generation, and…
- Qwen3 Coder Next48 GB+48.3 GB @Q4
Alibaba · 79.7B
Qwen3 Coder Next is a 79.7-billion parameter code-specialized instruction-tuned model from Alibaba Cloud, the next generation of the Qwen Coder series.
- Phi 28 GB+2.6 GB @Q4
Microsoft · 2.8B
Microsoft Phi 2 is a 2.8-billion parameter language model from Microsoft Research that pioneered the concept of small but highly capable language models.
- Qwen2.5 Coder 3B Instruct8 GB+2.8 GB @Q4
Alibaba · 3.1B
Qwen2.5 Coder 3B Instruct is a 3.1B-parameter open language model from Alibaba in the Qwen 2.5 family.
- Qwen2.5 Coder 1.5B Instruct8 GB+1.7 GB @Q4
Alibaba · 1.5B
Qwen2.5 Coder 1.5B Instruct is a 1.5B-parameter open language model from Alibaba in the Qwen 2.5 family.
- Qwen3 Coder 30B A3B Instruct16 GB+19.4 GB @Q4
Alibaba · 30.5B
Qwen3 Coder 30B A3B Instruct is a code-specialized Mixture of Experts (MoE) model from Alibaba Cloud's Qwen 3 Coder series, with 30 billion total parameters…
- Qwen2.5 Coder 1.5B8 GB+1.7 GB @Q4
Alibaba · 1.5B
Qwen2.5 Coder 1.5B is a 1.5-billion parameter code-specialized model from Alibaba Cloud's Qwen 2.5 Coder series.
- CodeLlama 7B HF12 GB+12.9 GB @Q4
Meta · 6.7B
CodeLlama 7B HF is a 6.7B-parameter open language model from Meta in the Code Llama family.
- Phi 4 Mini Instruct8 GB+4.8 GB @Q4
Microsoft · 3.8B
Microsoft Phi 4 Mini Instruct is a 3.8-billion parameter instruction-tuned model from Microsoft Research's Phi 4 family.
- Phi 3.5 Mini Instruct12 GB+9.0 GB @Q4
Microsoft · 3.8B
Phi 3.5 Mini Instruct is a 3.8B-parameter open language model from Microsoft in the Phi 3 family.
- Phi 3 Mini 4k Instruct8 GB+4.2 GB @Q4
Microsoft · 3.8B
Microsoft Phi 3 Mini 4K Instruct is a 3.8-billion parameter instruction-tuned model from Microsoft Research's Phi 3 generation, with a 4K token context window.
- Phi 412 GB+12.5 GB @Q4
Microsoft · 14.7B
Microsoft Phi 4 is a 14-billion parameter language model from Microsoft Research's Phi series, designed to deliver strong reasoning, mathematical, and coding performance at an…
- Qwen2.5 Coder 7B8 GB+5.8 GB @Q4
Alibaba · 7.6B
Qwen2.5 Coder 7B is a 7.6-billion parameter code-specialized base (pretrained) model from Alibaba Cloud's Qwen 2.5 Coder series.
- Xflux Text Encoders16 GB+5.4 GB @Q4
XLabs-AI · 4.8B
Xflux Text Encoders is a 4.8B-parameter open language model from XLabs-AI.
- Granite Docling 258M8 GB+0.6 GB @Q4
IBM · 258M
Granite Docling 258M is IBM Research's multimodal model with 258 million parameters, designed for document conversion.
- Taardis 27B Full Ternary21.4 GB @Q4
CodeMasterCody3D · 27B
Taardis 27B Full Ternary is a 27B-parameter open language model from CodeMasterCody3D.
- Phi 3 Mini 128k Instruct12 GB+9.0 GB @Q4
Microsoft · 3.8B
Phi 3 Mini 128k Instruct is a 3.8B-parameter open language model from Microsoft in the Phi 3 family.
- Tiny Starcoder Py8 GB+0.5 GB @Q4
BigCode · 164M
Tiny Starcoder Py is a 164M-parameter open language model from BigCode in the StarCoder family.
- Qwen2.5 Coder 0.5B Instruct8 GB+0.8 GB @Q4
Alibaba · 494M
Qwen2.5 Coder 0.5B Instruct is a 494M-parameter open language model from Alibaba in the Qwen 2.5 family.
- Deepseek Coder 6.7B Base12 GB+12.9 GB @Q4
DeepSeek · 6.7B
Deepseek Coder 6.7B Base is a 6.7B-parameter open language model from DeepSeek in the DeepSeek Coder family.
- Kimi K2.7 Code115617.6 GB @Q4
Moonshot AI · 1026.9B
Kimi-K2.7-Code is Moonshot AI's coding-focused agentic model, built on top of Kimi K2.6 with improvements aimed at long-horizon software engineering workflows and reduced reasoning-token usage.
- Phi 3.5 MoE Instruct24 GB+27.6 GB @Q4
Microsoft · 41.9B
Phi-3.5-MoE-instruct is Microsoft's Mixture-of-Experts model in the Phi-3.5 line, combining 16 experts of 3.8 billion parameters each, about 42 billion total, with only around 6.6…
- Sweep Next Edit 1.5B8 GB+1.6 GB @Q4
Xenova · 1.4B
Sweep Next Edit 1.5B is a 1.4B-parameter open language model from Xenova.
- Phi 1 58 GB+0.9 GB @Q4
Microsoft · 1.4B
Phi 1 5 is a 1.4B-parameter open language model from Microsoft in the Phi family.
- Sqlcoder 7B 216 GB+12.9 GB @Q4
defog · 6.7B
SQLCoder 7B 2 is a 6.7-billion-parameter model from Defog, purpose-built for converting natural-language questions into SQL queries.
- Codegemma 7B IT8 GB+8.2 GB @Q4
Google · 8.5B
Codegemma 7B IT is a 8.5B-parameter open language model from Google in the Gemma family.
- Qwen2.5 Coder 14B12 GB+12.4 GB @Q4
Alibaba · 14.8B
Qwen2.5 Coder 14B is a 14.8B-parameter open language model from Alibaba in the Qwen 2.5 family.
- GPT Bigcode Santacoder8 GB+0.7 GB @Q4
BigCode · 1.1B
GPT Bigcode Santacoder is a 1.1B-parameter open language model from BigCode.
- Phi 3 Vision 128k Instruct16 GB+4.9 GB @Q4
Microsoft · 4.1B
Phi 3 Vision 128k Instruct is a 4.1B-parameter open language model from Microsoft in the Phi 3 family.
- Phi 4 Mini Reasoning8 GB+4.8 GB @Q4
Microsoft · 3.8B
Phi-4-mini-reasoning is Microsoft's compact 3.8-billion-parameter model in the Phi-4 family, fine-tuned specifically for multi-step, logic-intensive mathematical reasoning using synthetic math data distilled from a larger…
- Codeguard Jev Style 1.5B8 GB+1.7 GB @Q4
JackKozmo29 · 1.5B
Codeguard Jev Style 1.5B is a 1.5B-parameter open language model from JackKozmo29.
- Granite 3B Code Base 2k8 GB+3.1 GB @Q4
IBM · 3.5B
Granite 3B Code Base 2k is a 3.5B-parameter open language model from IBM in the Granite family.
- CodeLlama 7B Instruct HF12 GB+12.9 GB @Q4
Meta · 6.7B
CodeLlama 7B Instruct HF is a 6.7B-parameter open language model from Meta in the Code Llama family.
- Qwen3 Coder 480B A35B Instruct101290.5 GB @Q4
Alibaba · 480.2B
Qwen3 Coder 480B A35B Instruct is Alibaba's largest code-specialized model, a massive 480.2-billion-parameter mixture-of-experts system with roughly 35 billion parameters active per token.
- Deepseek Coder 6.7B Instruct12 GB+12.9 GB @Q4
DeepSeek · 6.7B
DeepSeek Coder 6.7B Instruct is a first-generation code-specialized model trained on a large corpus of source code and programming-related data.
- Qwen2.5 Coder 0.5B8 GB+0.8 GB @Q4
Alibaba · 494M
Qwen2.5 Coder 0.5B is a 494-million parameter code-specialized model from Alibaba Cloud, the smallest in the Qwen 2.5 Coder series.
- Deepseek Coder 1.3B Instruct8 GB+4.3 GB @Q4
DeepSeek · 1.3B
DeepSeek Coder 1.3B Instruct is an ultra-compact code model designed for environments where hardware resources are extremely limited.
- Qwen3.5 4B Super Coder8 GB+3.0 GB @Q4
jica98 · 4B
Qwen3.5 4B Super Coder is a 4B-parameter open language model from jica98 in the Qwen 3.5 family.
- Starcoder2 3B8 GB+2.6 GB @Q4
BigCode · 3.0B
StarCoder2-3B is BigCode's 3-billion-parameter base code model, trained from scratch on 17 programming languages drawn from The Stack v2 using a fill-in-the-middle objective over more…
- Qwen3.8 Distill 35B A3B Coder Abliterated27.0 GB @Q4
Lord-H4D3ZS · 35B
Qwen3.8 Distill 35B A3B Coder Abliterated is a 35B-parameter open language model from Lord-H4D3ZS in the Qwen 3.8 family.
- Qwen2.5 Coder 3B8 GB+2.8 GB @Q4
Alibaba · 3.1B
Qwen2.5-Coder-3B is Alibaba's 3.1-billion-parameter base (pretrained, not instruction-tuned) code language model, one of six sizes in the Qwen2.5-Coder family built on Qwen2.5-3B.
- Codegen 350M Mono8 GB+0.2 GB @Q4
Salesforce · 350M
Codegen 350M Mono is a 350M-parameter open language model from Salesforce.
- Mythos Nano8 GB+2.8 GB @Q4
squ11z1 · 3.1B
Mythos Nano is a 3.1B-parameter open language model from squ11z1.
- Codegemma 2B8 GB+3.5 GB @Q4
Google · 2.5B
Codegemma 2B is a 2.5B-parameter open language model from Google in the Gemma 2 family.
- GLM 4.7 Flash Coder16 GB+19.1 GB @Q4
whitecircle · 29.9B
GLM 4.7 Flash Coder is a 29.9B-parameter open language model from whitecircle in the GLM 4 family.
- CodeLlama 34B HF24 GB+23.8 GB @Q4
Meta · 33.7B
CodeLlama 34B HF is a 33.7B-parameter open language model from Meta in the Code Llama family.
- MiMo V2.6 Distill Qwen 9B8 GB+6.5 GB @Q4
Xiaomi · 9.4B
MiMo V2.6 Distill Qwen 9B is Xiaomi's 9.4-billion-parameter agentic model, built by fine-tuning Alibaba's Qwen3.5-9B on data generated by Xiaomi's larger MiMo-V2.6-Pro and MiMo-V2.6-Flash models.
- Phi 4 Reasoning12 GB+12.5 GB @Q4
Microsoft · 14.7B
Phi-4-reasoning is Microsoft's reasoning-focused fine-tune of Phi-4, a 14-billion-parameter dense decoder-only transformer, produced through supervised fine-tuning on chain-of-thought traces followed by reinforcement learning, concentrated on…
- Phi 4 Reasoning Plus12 GB+12.5 GB @Q4
Microsoft · 14.7B
Phi 4 Reasoning Plus is a 14.7B-parameter open language model from Microsoft in the Phi 4 family.
- Huginn 012512 GB+4.7 GB @Q4
tomg-group-umd · 3.9B
Huginn 0125 is a 3.9B-parameter open language model from tomg-group-umd.
- Starcoder12 GB+13.6 GB @Q4
BigCode · 15.8B
Starcoder is a 15.8B-parameter open language model from BigCode in the StarCoder family.
- KAT Coder V2.5 Dev 35B A3B ABLITERATED UNCENSORED PHILADELPHIA CLASS24 GB+21.3 GB @Q4
KridgeDookie · 34.7B
KAT Coder V2.5 Dev 35B A3B ABLITERATED UNCENSORED PHILADELPHIA CLASS is a 34.7B-parameter open language model from KridgeDookie in the Phi family.
- MiMo V2.5 Pro116616.0 GB @Q4
Xiaomi · 1023.2B
MiMo V2.5 Pro is a 1023.2B-parameter open language model from Xiaomi.
- Deepseek Coder 1.3B Base8 GB+4.3 GB @Q4
DeepSeek · 1.3B
Deepseek Coder 1.3B Base is a 1.3B-parameter open language model from DeepSeek in the DeepSeek Coder family.
- CodeLlama 34B Instruct HF24 GB+23.8 GB @Q4
Meta · 33.7B
CodeLlama-34b-Instruct-hf is Meta's 34-billion-parameter instruction-tuned Code Llama model, fine-tuned from the base Code Llama checkpoint for safer, more reliable code-assistant use, as opposed to the…
- LocoOperator 4B8 GB+4.2 GB @Q4
LocoreMind · 4.0B
LocoOperator 4B is a 4.0B-parameter open language model from LocoreMind.
- Bloom122.4 GB @Q4
BigScience · 176.2B
Bloom is a 176.2B-parameter open language model from BigScience.
- Phi 3 Small 8k Instruct24 GB+6.0 GB @Q4
Microsoft · 7.4B
Phi-3-Small-8K-Instruct is Microsoft's 7-billion-parameter instruction-tuned chat model from the Phi-3 family, trained on a mix of synthetic data and heavily filtered public web text chosen…
- Qwopus3.6 27B Coder22.0 GB @Q4
Jackrong · 27.8B
Qwopus3.6 27B Coder is a 27.8B-parameter open language model from Jackrong.
- Qwen2.5 Coder 7B Instruct Abliterated8 GB+5.8 GB @Q4
huihui-ai · 7.6B
Qwen2.5 Coder 7B Instruct Abliterated is a 7.6B-parameter open language model from huihui-ai in the Qwen 2.5 family.
- DeepSeek Coder v2 Lite Base12 GB+10.2 GB @Q4
DeepSeek · 15.7B
DeepSeek Coder v2 Lite Base is a 15.7B-parameter open language model from DeepSeek in the DeepSeek Coder family.
- Phi 3 Medium 4k Instruct8 GB+9.5 GB @Q4
Microsoft · 14.0B
Phi 3 Medium 4k Instruct is a 14.0B-parameter open language model from Microsoft in the Phi 3 family.
- Jack 3.8 27B Coder 16GB VRAM21.4 GB @Q4
JackAgentLead · 27B
Jack 3.8 27B Coder 16GB VRAM is a 27B-parameter open language model from JackAgentLead.
- Phi 3 Medium 128k Instruct12 GB+12.0 GB @Q4
Microsoft · 14.0B
Phi-3-Medium-128K-Instruct is Microsoft's 14-billion-parameter instruction-tuned model in the Phi-3 family, trained on a mix of synthetic data and filtered high-quality web content chosen for reasoning…
- Phi 18 GB+0.9 GB @Q4
Microsoft · 1.4B
Phi 1 is a 1.4B-parameter open language model from Microsoft in the Phi family.
- Yi Coder 9B Chat8 GB+7.2 GB @Q4
01.AI · 8.8B
Yi-Coder 9B Chat is 01.AI's 9-billion-parameter instruction-tuned model for code, built upon the Yi-Coder-9B base checkpoint.
- KAT Coder V2.5 Dev24 GB+21.3 GB @Q4
Kwaipilot · 34.7B
KAT Coder V2.5 Dev is a 34.7B-parameter open language model from Kwaipilot.
- VibeThinker 3B8 GB+2.8 GB @Q4
WeiboAI · 3.1B
VibeThinker 3B is a 3.1B-parameter open language model from WeiboAI.
- IQuest Coder V1 40B Loop Instruct30.3 GB @Q4
IQuestLab · 39.8B
IQuest Coder V1 40B Loop Instruct is a 39.8B-parameter open language model from IQuestLab.
- Qwen2.5 Coder 3B Claude Opus 4.6 Distilled8 GB+2.8 GB @Q4
ryzdfm · 3.1B
Qwen2.5 Coder 3B Claude Opus 4.6 Distilled is a 3.1B-parameter open language model from ryzdfm in the Qwen 2.5 family.
- Starcoder2 7B8 GB+5.7 GB @Q4
BigCode · 7.2B
Starcoder2 7B is a 7.2B-parameter open language model from BigCode in the StarCoder family.
- Huihui Qwen3 Coder 30B A3B Instruct Abliterated16 GB+19.4 GB @Q4
huihui-ai · 30.5B
Huihui Qwen3 Coder 30B A3B Instruct Abliterated is a 30.5B-parameter open language model from huihui-ai in the Qwen 3 family.
- North Mini Code 1.016 GB+19.4 GB @Q4
Cohere · 30.5B
North Mini Code 1.0 is Cohere's first open-weight coding model, a mixture-of-experts design with about 30.5 billion total parameters and roughly 3.3 billion active per…
- DeepSeek Coder v2 Instruct142.9 GB @Q4
DeepSeek · 235.7B
DeepSeek-Coder-V2-Instruct is DeepSeek's mixture-of-experts code and math model, continuing pretraining from an intermediate DeepSeek-V2 checkpoint with an additional 6 trillion tokens focused on code.
- DeepHat V1 7B8 GB+5.8 GB @Q4
DeepHat · 7.6B
DeepHat V1 7B is a 7.6B-parameter open language model from DeepHat.
- Gemma 4 31B IT Heretic24 GB+21.7 GB @Q4
coder3101 · 31.3B
Gemma 4 31B IT Heretic is a 31.3B-parameter open language model from coder3101 in the Gemma 4 family.
- Starcoder2 15B12 GB+11.2 GB @Q4
BigCode · 16.0B
Starcoder2 15B is a 16.0B-parameter open language model from BigCode in the StarCoder family.
- Whittle Next 27B A3B16 GB+17.5 GB @Q4
logic65 · 27.6B
Whittle Next 27B A3B is a 27.6B-parameter open language model from logic65.
- Deepseek Coder 33B Instruct24 GB+24.5 GB @Q4
DeepSeek · 33.3B
DeepSeek-Coder-33B-Instruct is DeepSeek's 33.3-billion-parameter instruction-tuned code model, initialized from DeepSeek-Coder-33B-Base and further fine-tuned on 2 billion tokens of instruction data for chat-style code generation, completion…
- Qwen3.8 27B Escha W216 GB+6.6 GB @Q4
EschaLabs · 6.3B
Qwen3.8 27B Escha W2 is a 6.3B-parameter open language model from EschaLabs in the Qwen 3.8 family.
- Kimi Dev 72B48 GB+49.3 GB @Q4
Moonshot AI · 72.7B
Kimi Dev 72B is Moonshot AI's developer-focused model built on the Qwen2.5-72B architecture, specifically optimized for coding tasks, tool use, and agentic workflows.
- OxCoder 9B24 GB+8.9 GB @Q4
OrionLLM · 9.4B
OxCoder 9B is a 9.4B-parameter open language model from OrionLLM.
- Qwen3.8 Flash Coder 85gb BF1632.2 GB @Q4
Jab1718 · 42.6B
Qwen3.8 Flash Coder 85gb BF16 is a 42.6B-parameter open language model from Jab1718 in the Qwen 3.8 family.
- UltiMerge26.8 GB @Q4
OliviaRossi · 34.7B
UltiMerge is a 34.7B-parameter open language model from OliviaRossi.
- Mellum 4B Base8 GB+5.7 GB @Q4
JetBrains · 4.0B
Mellum 4B Base is a 4.0B-parameter open language model from JetBrains in the Mellum family.
- Qwen3.8 27B Pi16 GB+17.9 GB @Q4
bytkim · 27.8B
Qwen3.8 27B Pi is a 27.8B-parameter open language model from bytkim in the Qwen 3.8 family.
- Moderato V1 Pro80.2 GB @Q4
nitrai-research · 113.3B
Moderato V1 Pro is a 113.3B-parameter open language model from nitrai-research.
- Qwen2.5 Coder 7B Bird Cot8 GB+5.8 GB @Q4
jk200201 · 7.6B
Qwen2.5 Coder 7B Bird Cot is a 7.6B-parameter open language model from jk200201 in the Qwen 2.5 family.
- Qwen3 42B A3B 2507 Thinking Abliterated Uncensored TOTAL RECALL v2 Medium MASTER CODER24 GB+26.8 GB @Q4
DavidAU · 42.4B
Qwen3 42B A3B 2507 Thinking Abliterated Uncensored TOTAL RECALL v2 Medium MASTER CODER is a 42.4B-parameter open language model from DavidAU in the Qwen 3…
- NEXUS Coder8 GB+1.7 GB @Q4
fableforge-ai · 1.5B
NEXUS Coder is a 1.5B-parameter open language model from fableforge-ai.
- Qwen2.5 Coder 32B24 GB+24.3 GB @Q4
Alibaba · 32.8B
Qwen2.5-Coder-32B is Alibaba's 32.8-billion-parameter code-specialized base language model, the largest of six sizes in the Qwen2.5-Coder family (0.5B to 32B), continued-pretrained from Qwen2.5-32B on 5.5…
- Instinct Python Coder Gemma4 12B Qwen3.8 27B12 GB+10.8 GB @Q4
projectj · 12B
Instinct Python Coder Gemma4 12B Qwen3.8 27B is a 12B-parameter open language model from projectj in the Qwen 3.8 family.
- Huihui Qwen3 Coder Next Abliterated48 GB+48.3 GB @Q4
huihui-ai · 79.7B
Huihui Qwen3 Coder Next Abliterated is a 79.7B-parameter open language model from huihui-ai in the Qwen 3 family.
- Gemma 4 E2B IT Qat Q4 0 Unquantized Heretic8 GB+3.5 GB @Q4
coder3101 · 5.1B
Gemma 4 E2B IT Qat Q4 0 Unquantized Heretic is a 5.1B-parameter open language model from coder3101 in the Gemma 4 family.
- Llama 3.2 1B MathCodeInstruct 5k8 GB+1.6 GB @Q4
OliverSundaram · 1.2B
Llama 3.2 1B MathCodeInstruct 5k is a 1.2B-parameter open language model from OliverSundaram in the Llama 3 family.
- Qwopus3.5 4B Coder12 GB+5.3 GB @Q4
Jackrong · 4.7B
Qwopus3.5 4B Coder is a 4.7B-parameter open language model from Jackrong.
- Deepseek Coder 33B Base24 GB+24.5 GB @Q4
DeepSeek · 33.3B
DeepSeek-Coder-33B-Base is DeepSeek's 33.3-billion-parameter base (pretrained, not instruction-tuned) code model, trained from scratch on 2 trillion tokens (87% code, 13% natural language in English and…
- Kimi K2.7 Code DFlash8 GB+2.7 GB @Q4
NVIDIA · 3.5B
Kimi K2.7 Code DFlash is a 3.5B-parameter open language model from NVIDIA in the Kimi K2 family.
- Jan Code 4B8 GB+4.5 GB @Q4
janhq · 4.4B
Jan Code 4B is a 4.4B-parameter open language model from janhq.
- LocoTrainer 4B8 GB+4.2 GB @Q4
LocoreMind · 4.0B
LocoTrainer 4B is a 4.0B-parameter open language model from LocoreMind.
- Fable Coder 35B A3B27.6 GB @Q4
Achilles1089 · 36.0B
Fable Coder 35B A3B is a 36.0B-parameter open language model from Achilles1089.
- ALoDLM 8B8 GB+7.6 GB @Q4
Amazon · 8.2B
ALoDLM-8B is Amazon's 8-billion-parameter diffusion language model built on a Qwen3 backbone.
- Soren 1 Small8 GB+3.0 GB @Q4
syntropy-ai · 1.9B
Soren 1 Small is a 1.9B-parameter open language model from syntropy-ai.
- CodeQwen1.5 7B8 GB+5.7 GB @Q4
Alibaba · 7.3B
CodeQwen1.5-7B is Alibaba's 7.3-billion-parameter code-specialized base language model, built on the Qwen1.5 architecture and pretrained on 3 trillion tokens of code data covering 92 programming…
- Phi 4 Mini Flash Reasoning8 GB+5.3 GB @Q4
Microsoft · 3.9B
Phi 4 Mini Flash Reasoning is a 3.9B-parameter open language model from Microsoft in the Phi 4 family.
- Ornith 1.5 35B A3B REAP 50 NVFP4A1648 GB+15.5 GB @Q4
Ttimms · 18.5B
Ornith 1.5 35B A3B REAP 50 NVFP4A16 is a 18.5B-parameter open language model from Ttimms in the Ornith family.
- Sweep Next Edit v2 7B8 GB+5.8 GB @Q4
sweepai · 7.6B
Sweep Next Edit v2 7B is a 7.6B-parameter open language model from sweepai.
- VibeThinker 1.5B8 GB+1.8 GB @Q4
WeiboAI · 1.8B
VibeThinker 1.5B is a 1.8B-parameter open language model from WeiboAI.
- MiMo V2.5 Pro Base616.0 GB @Q4
Xiaomi · 1023.2B
MiMo V2.5 Pro Base is a 1023.2B-parameter open language model from Xiaomi.
- CodeLlama 70B Instruct HF48 GB+43.0 GB @Q4
Meta · 69.0B
CodeLlama-70B-Instruct is Meta's largest Code Llama model, a 69-billion-parameter, instruction-tuned member of the Code Llama family (7B to 70B) built on Llama 2 for general…
- OmniCoder 9B24 GB+8.9 GB @Q4
Tesslate · 9.4B
OmniCoder 9B is a 9.4B-parameter open language model from Tesslate.
- OpenReasoning Nemotron 32B24 GB+24.3 GB @Q4
NVIDIA · 32.8B
OpenReasoning Nemotron 32B is a 32.8B-parameter open language model from NVIDIA in the Nemotron family.
- Stable DiffCoder 8B Instruct24 GB+6.6 GB @Q4
ByteDance-Seed · 8.3B
Stable DiffCoder 8B Instruct is a 8.3B-parameter open language model from ByteDance-Seed.
- AesCode 32B24 GB+23 GB @Q4
Microsoft · 33.4B
AesCode-32B is a Microsoft research model that writes slides, posters, dashboards, and infographics as editable HTML and CSS.
- AesCode 8B8 GB+8.0 GB @Q4
Microsoft · 8.8B
AesCode-8B is the smaller of Microsoft's two AesCode models, which generate slides, posters, dashboards, and infographics as structured, editable HTML and CSS.
- Steelman 14B Ada12 GB+12.2 GB @Q4
the-clanker-lover · 14B
Steelman 14B Ada is a 14B-parameter open language model from the-clanker-lover.
- NousCoder 14B12 GB+11.8 GB @Q4
Nous Research · 14.8B
NousCoder 14B is a 14.8B-parameter open language model from Nous Research.
- Jan v3 4B Base Instruct8 GB+4.5 GB @Q4
janhq · 4.4B
Jan v3 4B Base Instruct is a 4.4B-parameter open language model from janhq.
- Kai 30B Instruct24 GB+24.3 GB @Q4
NoesisLab · 32.8B
Kai 30B Instruct is a 32.8B-parameter open language model from NoesisLab.
- Beam 1.0 501B A23B304.0 GB @Q4
reflection · 501.3B
Beam 1.0 is an open model from Reflection, a sparse mixture-of-experts language model with about 501 billion total parameters and 23 billion active per token.
- Phi 4 Quantized.w8a812 GB+12.5 GB @Q4
RedHatAI · 14.7B
Phi 4 Quantized.w8a8 is a 14.7B-parameter open language model from RedHatAI in the Phi 4 family.
- Dhara 70M8 GB+0.1 GB @Q4
codelion · 71M
Dhara 70M is a 71M-parameter open language model from codelion.
- Nerdsking Python Coder 7B I8 GB+7.1 GB @Q4
Nerdsking · 7B
Nerdsking Python Coder 7B I is a 7B-parameter open language model from Nerdsking.
- Qwen3 Code Reasoning 4B8 GB+4.2 GB @Q4
GetSoloTech · 4B
Qwen3 Code Reasoning 4B is a 4B-parameter open language model from GetSoloTech in the Qwen 3 family.
- SmolLM2 70M8 GB+0.5 GB @Q4
codelion · 69M
SmolLM2 70M is a 69M-parameter open language model from codelion in the SmolLM family.
- Mistral Small 3.2 24B Qiskit16 GB+18.1 GB @Q4
Qiskit · 24.0B
Mistral Small 3.2 24B Qiskit is a 24.0B-parameter open language model from Qiskit in the Mistral family.
- OpenCodeReasoning Nemotron 1.1 32B24 GB+24.3 GB @Q4
NVIDIA · 32.8B
OpenCodeReasoning Nemotron 1.1 32B is a 32.8B-parameter open language model from NVIDIA in the Nemotron family.
- NextCoder 7B8 GB+5.8 GB @Q4
Microsoft · 7.6B
NextCoder 7B is a 7.6B-parameter open language model from Microsoft.
- XiYanSQL QwenCoder 32B 250424 GB+23.8 GB @Q4
XGenerationLab · 32B
XiYanSQL QwenCoder 32B 2504 is a 32B-parameter open language model from XGenerationLab in the Qwen family.
- Deepseek Coder 1.3B Kexer8 GB+4.3 GB @Q4
JetBrains · 1.3B
Deepseek Coder 1.3B Kexer is a 1.3B-parameter open language model from JetBrains in the DeepSeek Coder family.
How much VRAM do you need for a local coding model?
A coding model's hardware needs scale with parameter count and quantization. A 7B model runs comfortably on an 8 GB GPU at Q4_K_M; a 32B coder wants ~20 GB (a 24 GB card like the RTX 3090/4090); and mixture-of-experts coders like Qwen3-Coder-30B-A3B run far faster than their total size suggests because only a few billion parameters are active per token. Apple Silicon Macs with 32 GB+ of unified memory are excellent for larger coders. Open any model to see its full VRAM-by-quantization table and the exact hardware that fits.
Frequently Asked Questions
What is the best local LLM for coding in 2026?
For most setups, Qwen2.5-Coder (7B for 8 GB GPUs, 32B for 24 GB) and Qwen3-Coder-30B-A3B are the strongest open coding models you can run locally. Microsoft's Phi-4 is a great smaller all-rounder. The right pick depends on your VRAM — open any model below to confirm it fits your hardware.
Can I run a coding LLM on my laptop?
Yes. 1.5B–7B coding models (e.g. Qwen2.5-Coder 1.5B/7B) run on laptops with 8–16 GB of RAM or VRAM via Ollama or LM Studio. MoE models with low active-parameter counts also run well on modest hardware.
Are local coding models as good as cloud models?
The best open coders (32B+ and large MoE) now rival mid-tier cloud models for everyday coding, autocomplete, and refactoring, while running fully offline and privately. Very large frontier models still lead on the hardest tasks, but the gap keeps closing.
How do I run these coding models?
Install Ollama or LM Studio, then pull the model — for example `ollama run qwen2.5-coder:7b`. Each model page lists the exact install command and the GPUs and Macs that can run it at each quantization level.