Best LLMs for coding: developer favorites

[Last updated: May 2025]

The best LLMs that developers use for coding stand out by combining deep understanding of programming languages with practical capabilities that enhance a developer's workflow. They solve complex problems and deliver code that can be used to build production applications faster – not just vibe code a prototype. 

These models don't just generate syntactically correct code, but understand context, purpose, and best practices across various languages, frameworks, and libraries.

Many of these LLMs are available to use in developer tools like Cursor, Codex, and GitHub Copilot. Software developers tend to have a favorite LLM for code completion and use a few different models depending on the specific task. Here are some of the LLMs developers use the most for coding.

Developers’ favorite LLMs for coding

There may be a new coding LLM leading the pack by the time you read this, but there are some consistent favorites software engineers rely on. 

Claude 4

Claude Opus 4 - Claude Opus 4 excels at coding and complex problem-solving, powering frontier AI agents. Cursor calls it state-of-the-art for coding and a leap forward in complex codebase understanding. It’s designed for sustained performance on long-running tasks that require focused effort and thousands of steps, with the ability to work continuously for several hours.

Claude Sonnet 4 - A significant upgrade delivering superior coding and reasoning while responding more precisely to instructions. Achieves state-of-the-art score of 72.7% on SWE-bench – balancing performance and efficiency for both internal and external use cases. GitHub says Claude Sonnet 4 soars in agentic tasks and will introduce it as the base model for the new coding agent in GitHub Copilot.

Key capabilities:

• Extended thinking with tool use (beta): Both models can use tools – like web search – during extended thinking, allowing Claude to alternate between reasoning and tool use to improve responses
• World-class coding performance: Claude Opus 4 achieves 72.5% on SWE-bench (79.4% with parallel test-time compute), while Sonnet 4 reaches 72.7% on SWE-bench Verified (80.2% with parallel test-time compute).
• API access: Available via API with four new Agent capabilities – the code execution tool, MCP connector, Files API, and the ability to cache prompts for up to one hour.
• Parallel tool execution: Both models can use multiple tools simultaneously for enhanced productivity
• Enhanced memory capabilities: When given access to local files, demonstrates significantly improved memory capabilities, extracting and saving key facts to maintain continuity and build tacit knowledge over time
• Hybrid reasoning modes: Offer both near-instant responses and extended thinking for deeper reasoning
• Reduced shortcut behavior: 65% less likely to engage in shortcuts or loopholes compared to Sonnet 3.7 on agentic tasks
• Integration with Claude Code: Works seamlessly with Anthropic's now generally available agentic development tool
• Sustained focus: Opus 4 delivers sustained performance on complex, long-running tasks and agent workflows
• Advanced instruction following: Both models follow instructions more precisely and demonstrate enhanced steerability

Industry feedback highlights Claude Opus 4 as "state-of-the-art for coding" with dramatic improvements in complex codebase understanding, while Sonnet 4 is being adopted as the base model for GitHub Copilot's new coding agent.

Claude 3.7 Sonnet

Claude is a consistent favorite for advanced coding tasks and reliable enterprise tools. With its exceptional reasoning capabilities and impressive 62.3% accuracy on SWE-bench (70.3% with custom scaffolding), it has strong traction among professional developers.

Key capabilities:

• State-of-the-art performance: Achieves 70.3% accuracy on SWE-bench Verified in standard mode, leading the industry in real-world software engineering tasks
• Hybrid reasoning: Can toggle between quick responses and extended step-by-step thinking that's visible to users
• Full development lifecycle: Excels across planning, coding, debugging, and maintenance stages
• Extensive output capacity: Supports up to 128K output tokens (over 15x longer than predecessors), crucial for complex code generation
• Strong in multimodal tasks: Exceptional at understanding code in context of visual elements (like diagrams or screenshots)
• System architecture insights: Particularly strong at planning large refactors and suggesting architectural improvements
• Advanced pattern recognition: Quickly identifies recurring code patterns and suggests optimizations
• Precision in code completion: Generates code that matches the style and conventions of existing projects
• Complex project navigation: Effectively handles large codebases and manages dependencies between components

According to Cursor's extensive testing, Claude 3.7 Sonnet is "best-in-class for real-world coding tasks" with significant improvements in handling complex codebases and advanced tool use.

GPT-4.1, GPT-4o, and o3/o4 models

OpenAI's latest LLMs are ideal for planning, code completion, and perform the best across multiple benchmarks. They remain the standard that other models are measured against. The newer o3 series is particularly favored for STEM and technical applications.

Key capabilities:

• GPT-4.1: Offers improved efficiency over GPT-4.5 while maintaining strong performance; particularly good at code optimization and security analysis
• GPT-4o: Excels at multimodal tasks, understanding code alongside diagrams, screenshots, and other visual elements; strong at rapid prototyping
• o3 Series: Specifically designed for reasoning-intensive tasks; particularly powerful for algorithmic optimization and mathematically complex code
• o3-mini: More affordable version of o3 with 49.3% accuracy on SWE-bench; excellent for educational settings and smaller development teams
• Robust documentation generation: Creates comprehensive and well-structured documentation for complex codebases
• Test coverage generation: Identifies edge cases and generates exhaustive test suites automatically
• Framework adaptation: Quickly adapts to different frameworks and libraries, understanding their patterns and conventions
• Security vulnerability identification: Identifies potential security issues and suggests mitigation strategies
• Refactoring excellence: Particularly strong at suggesting and implementing complex refactorings

The o3 series is particularly favored for STEM and technical applications requiring rigorous logical reasoning, though it comes at a higher cost than some alternatives.

Gemini 2.5 Pro w/ Deep Think

Google's Gemini 2.5 Pro has made enormous strides in 2025, now leading the WebDev Arena coding leaderboard and showing particular strength in full-stack development. While Claude 3.7 Sonnet has had a steady lead, many developers are excited about using Gemini LLMs for coding.

• Deep Think reasoning mode: Considers multiple hypotheses before responding, leading to superior results on complex tasks
• Top performance on LiveCodeBench: Leading position on this difficult benchmark for competition-level coding
• Exceptional multimodal understanding: 84% score on MMMU for multimodal reasoning, particularly valuable for visual programming contexts
• Full-stack expertise: Balanced capabilities across frontend and backend development in multiple programming languages
• Thought summarization: Provides organized, clear explanations of reasoning process with headers and key details
• Configurable thinking budgets: Allows developers to control the balance between response quality and latency
• Project Mariner integration: Computer use capabilities for direct interaction with development environments
• Enhanced security awareness: Strong protections against prompt injection and security vulnerabilities
• Superior database query optimization: Particularly strong at generating and optimizing complex database queries
• Cross-language translation: Excels at converting code between different programming languages

Gemini 2.5 Pro consistently gets compared to Claude 3.7 and OpenAI's latest LLMs on Reddit, with developers often praising its well-balanced full-stack capabilities.

DeepSeek R1 & V3

DeepSeek R1 shocked the AI community in 2025 by demonstrating competitive performance against leading frontier models at a fraction of the cost.

Key capabilities:

• Mixture of Experts architecture: 671 billion parameter MoE model with 37 billion activated parameters per token
• Mathematical prowess: Exceptional at mathematically intensive algorithms with 97.3% on MATH-500 benchmark
• SWE-bench performance: 49.2% on SWE-bench Verified, slightly ahead of OpenAI's o1-1217 at 48.9%
• Advanced code generation: 96.3% percentile ranking on Codeforces, demonstrating strong algorithmic reasoning
• Cost efficiency: Approximately 15-50% of the cost of models like OpenAI's o1
• Open-source flexibility: MIT license allowing commercial use and modifications
• Distilled variants: Smaller distilled models based on Llama and Qwen for specific use cases
• Specialized reasoning: DeepSeek R1 focuses on advanced reasoning while DeepSeek V3 is better for general coding tasks
• Transparent thinking process: Reasoning steps are visible to users, allowing for better understanding and debugging

While DeepSeek can perform some advanced coding tasks, many senior engineers report using it as a secondary model, often leveraging its cost advantages for specific types of problems.

Multi-model approach to coding with LLMs

Many developers use multiple models through platforms like Cursor or specialized IDE integrations like Codex CLI – leveraging different models' strengths for specific coding tasks rather than relying on a single LLM for every task.

• Code completion and FIM: Claude 3.7 Sonnet provides excellent code completion with its advanced pattern recognition, adapting precisely to existing code styles and maintaining consistency. OpenAI’s most advanced LLMs are also often used as a go-to for code completion.
• Architecture and design: Claude 3.7 Sonnet excels at system design, planning large refactors, and maintaining consistency across complex codebases
• Algorithm development: DeepSeek R1 is particularly strong for mathematically intensive algorithms and optimization problems
• Security auditing: OpenAI's GPT-4.1 and o3 series demonstrate superior capabilities in identifying security vulnerabilities
• Documentation generation: Gemini 2.5 Pro with its thought summarization features creates exceptionally clear and comprehensive documentation
• Multi-language projects: Models with strong cross-language capabilities like GPT-4o help with projects spanning multiple programming languages
• Educational settings: More affordable options like o3-mini or DeepSeek's distilled models provide excellent learning tools

Aside from developer’s preferences, you can get an idea of how good an LLM is at coding with industry benchmarks. 

Coding LLM benchmarks

These are the leading benchmarks for evaluating LLMs in software development – from Python-specific tasks to real-world software engineering tasks. 

1. HumanEval - Developed by OpenAI, this benchmark evaluates how effectively LLMs generate code through 164 Python programming problems with unit tests.
   
   
2. SWE-bench - The industry standard for agentic code evaluation, measuring how models perform on real-world software engineering tasks.
   
   
3. DevQualityEval - Focuses on assessing models' ability to generate quality code across multiple languages including Java, Go, and Ruby.
   
   
4. LiveCodeBench - A holistic and contamination-free evaluation benchmark that continuously collects new problems over time, focusing on broader code-related capabilities.
   
   
5. Aider-Polyglot - Tests models' abilities to edit source files to complete coding exercises across multiple programming languages.
   
   
6. BigCodeBench - The next generation of HumanEval, focusing on more complex function-level code generation with practical challenging coding tasks.
   
   
7. Chatbot Arena - Provides head-to-head comparisons of various models on coding and general tasks.
   
   
8. MBPP - Mostly Basic Python Problems benchmark assesses code generation through more than 900 coding tasks.

These benchmarks help developers compare key metrics such as debugging accuracy, code completion capabilities, and performance on complex programming challenges.

Which LLM should my team use for coding?

Different models excel in different areas:

• Claude 3.7 Sonnet stands out for exceptional reasoning and software engineering accuracy
• OpenAI's LLMs provide strong performance across diverse technical challenges
• Gemini 2.5 Pro with Deep Think offers impressive full-stack capabilities
• DeepSeek delivers competitive performance at a more accessible price point

Most experienced developers take a multi-model approach –  leveraging specialized strengths for specific tasks through platforms like Cursor and Copilot.

The ideal LLM depends on your specific needs:

• Code completion and maintenance
• Architecture design
• Algorithm development
• Documentation generation
• Code refactoring

Ultimately, experiment with several options to find the combination that best enhances your development workflow and delivers the quality results your projects demand.

Need help choosing which coding LLMs to use at your company?

We can help you decide which coding LLMs are best for your teams, tech stack, and goals. Our senior engineering teams already use AI coding tools to build production-ready software.

And if you need to hire an AI-augmented team of senior software engineers to speed up delivery, we can assemble them for you in 4-6 weeks. It’ll be faster to get started, more cost-efficient than internal hiring, and we’ll deliver high-quality results quickly.

Zappos, Twilio, and Veho are just a few companies that trust us to build their software and enterprise systems.

You can schedule a time to talk with us here. No hassle, no expectations, just answers.