10 categories and 65 products and companies, charted by an ai agent.
General-purpose code LLMs are foundation models pretrained on large corpora of source code and natural language to enable broad coding tasks like generation, completion, and explanation. They are used by developers, platform builders, and enterprises integrating AI into IDEs, CI/CD pipelines, and developer tools.
gpt-4o · claude sonnet 4.5 · gemini 2.5 pro · deepseek-v3 · grok-3 · mistral large · llama 3.3 · qwen2.5-coder · o3
Specialized code base models are large language models pretrained or fine-tuned specifically on source code and programming languages to enable code generation, completion, and understanding. They are used by developers, DevOps teams, and enterprises seeking to automate software development workflows.
code llama · starcoder2 · deepseek-coder-v2 · codestral · incoder · granite code · phi-4 · starcoder · glm · minimax
IDE-integrated AI coding assistants embed code completion, generation, and analysis directly into development environments like VS Code and JetBrains IDEs. Software developers use them to write, refactor, and debug code within their existing workflows.
github copilot · cursor · windsurf · amazon q developer · codeium · tabnine · void · jetbrains ai · zed
Agentic SWE systems are AI-powered software engineering agents that autonomously write, debug, review, and iterate on code end-to-end with minimal human intervention. They are used by developers, engineering teams, and enterprises looking to accelerate software delivery and reduce manual coding burden.
devin · swe-agent · openai codex (agent) · aide · moatless tools · all-hands opendevin
RL-trained code reasoners are foundation models fine-tuned with reinforcement learning to solve complex programming tasks through multi-step reasoning and self-correction. They are used by developers, competitive programmers, and engineering teams seeking AI that can debug, generate, and optimize code with deeper logical rigor than standard LLMs.
deepseek-r1 · o3-mini · qwq-32b · kimi k1.5 · s1
Full-stack AI app builders provide code foundation models as core infrastructure for developers to construct applications. Data scientists, ML engineers, and application developers use these models to generate, complete, and reason about code across entire projects.
bolt · lovable · v0 · replit agent · val town · replit · base44
Systems that analyze code for defects, style violations, and security issues, then report findings to developers and development teams. Used by software engineers, QA professionals, and DevOps practitioners to catch problems before deployment.
coderabbit · qodo (codiumai) · sourcery · ellipsis
Foundation models trained specifically to generate, understand, and reason about code at scale. Used by developers, enterprises, and tool builders to power code completion, generation, and analysis workflows.
claude · openai codex · gemini · grok code · deepseek coder · qwen coder · kimi k2
Autonomous coding agents are systems that can plan, execute, and iterate on multi-step software engineering tasks with minimal human intervention. Data engineers, platform teams, and development organizations use them to automate code generation, bug fixes, refactoring, and task decomposition.
claude code · openhands · cline · factory · amp
Tools that automatically analyze code for defects, style violations, and security issues, then provide feedback or fixes. Used by developers and engineering teams during development and in CI/CD pipelines.
qodo · graphite · sourcegraph