Code reviews

Models 22

• Gen 7
  MiMo V2.5 Pro

  By Xiaomi

  MiMo-V2.5-Pro is Xiaomi’s latest agentic language model, built for stronger long-horizon coherence, complex software engineering, and autonomous tool-using workflows. It is positioned as Xiaomi’s most capable model so far, with better instruction following in agent settings, stronger ultra-long-context consistency, and support for very long tool-use trajectories.

  💻Coding🔍Code reviews✉️Codebase Q&A🤖Agents🎩Sophisticated reasoning
  NewText

  Released 1mo ago
• Gen 7
  SWE 1.6

  By Cognition AI

  SWE-1.6 is Cognition’s latest model for software engineering agents, built to improve both raw capability and what the company calls “model UX.” It is positioned as fast, low-cost, and smoother to use in practice, with up to 950 tok/s and better behavior in agent workflows.

  💻Coding🔍Code reviews🐞Debugging
  NewText

  Released 1mo ago
• Gen 7
  Trinity Large Thinking

  By Arcee AI

  Trinity-Large-Thinking is Arcee AI’s open-source reasoning model for long-horizon agents and multi-turn tool use. It is built to improve context coherence, instruction following, and stability across extended agent loops, and is released under Apache 2.0 for developers and enterprises that want open weights they can inspect, host, and post-train.

  💬Chatting🔍Code reviews💻Conversational coding🧠AI inference
  NewText

  Released 2mo ago
• Gen 3 Qwen
  Qwen 3.6 Plus

  By Alibaba

  Qwen3.6 Plus is Alibaba’s next-generation preview model in the Qwen Plus line, built with a hybrid architecture for better efficiency, scalability, reasoning, and agentic behavior. It is positioned as especially strong for coding, front-end development, and complex multi-step problem solving, with support for very long-context use cases.

  💬Chatting🌐Text translation🔍Code reviews💻Conversational coding
  NewMultimodal

  Released 2mo ago
• Gen 7
  Devin 2.2

  By Cognition AI

  Devin 2.2 is a major Devin update that adds stronger autonomy via computer use for end-to-end testing, plus self-verification and auto-fix loops.

  🧪Software testing🔍Code reviews🕸️Browser testing
  Text

  Released 3mo ago
• Gen 3
  Kilo

  By Kilo Code

  Kilo is an agentic engineering platform and open-source coding agent that can generate code from natural language, run terminal commands, and automate browser-based steps, available as a VS Code extension and a CLI.

  💻Coding🔍Code reviews✉️Codebase Q&A🔧Code refactoring👨‍💻Collaborative coding
  Multimodal

  Released 3mo ago
• Gen 2 GPT
  GPT 5.3 Codex

  By OpenAI

  GPT-5.3-Codex is OpenAI’s most capable agentic coding model, combining GPT-5.2-Codex’s frontier coding with GPT-5.2’s reasoning in one model that is ~25% faster, tops SWE-Bench Pro and Terminal-Bench, and runs long, tool-using workflows on a computer.

  💻Coding🔍Code reviews🔎Problem solving📝Writing
  Coding

  Released 3mo ago
• Gen 2 Qwen
  Qwen 3 Coder Next

  By Alibaba

  Qwen3-Coder-Next is an open-weight language model built specifically for coding agents and local development, optimized for tool use and long-horizon software engineering tasks.

  💻Coding🔍Code reviews🔧Code refactoring🎯Code autocompletion💻Software engineering guidance
  Coding

  Released 4mo ago
• Gen 7 SERA
  SERA 32B

  By Ai2

  SERA-32B is Ai2’s flagship 32B Open Coding Agent, a Qwen 3-32B based repository-level coding model trained with Soft Verified Generation that reaches about 49.5% on SWE-bench Verified, rivaling much larger proprietary systems.

  💻Coding📚Code explanations🔍Code reviews💡Coding help
  Text

  Released 4mo ago
• Gen 7 SERA
  SERA-8B

  By Ai2

  SERA-8B is Ai2’s 8B-parameter open-source coding agent, fine-tuned from Qwen 3-8B with Soft Verified Generation to solve repository-level bugs, reaching about 31.7% resolve rate on SWE-bench Verified at 32K context.

  💻Coding📚Code explanations🔍Code reviews
  Text

  Released 4mo ago
• Gen 7
  Devstral 2

  By Mistral AI

  Devstral 2 is Mistral AI’s 123B open-weight coding model family, paired with the open-source Mistral Vibe CLI agent, to explore, edit and execute changes across large codebases with a 256K context window for autonomous software engineering workflows

  💻Coding🔍Code reviews💡Coding help
  Text

  Released 5mo ago
• Gen 3 Grok
  Grok 4.1

  By xAI

  Grok 4.1 is xAI’s reasoning model tuned for stronger analysis and coding with faster, more direct answers. It supports long context, tool and function calling, streaming, and clean JSON, making it a solid default for assistants, RAG, and agent workflows.

  💻Coding🔍Code reviews🔎Problem solving🔬Code testing
  Multimodal

  Released 6mo ago
• Gen 2
  KAT Dev 32B

  By Kuaishou Technology

  KAT-Dev-32B is Kwaipilot’s open-source 32B software engineering model built for real-world coding tasks. It is trained with mid-training, supervised fine-tuning, reinforcement fine-tuning, and large-scale agentic RL, and is positioned as a strong open model for software development workflows such as bug fixing, feature work, refactoring, testing, and deployment.

  💻Coding🔍Code reviews🔧Code refactoring💻Software engineering guidance
  Multimodal

  Released 8mo ago
• Gen 7 Qwen
  Qwen 3 Omni Flash

  By Alibaba

  Qwen 3 Coder Flash is the speed-tuned coding variant. It reads repos, writes diffs and tests, and supports long context, tool calling, and strict JSON or patch outputs.

  💻Coding📚Code explanations🔍Code reviews
  Text

  Released 8mo ago
• Gen 7 Nemotron
  Nemotron Nano 2

  By NVIDIA

  Nemotron-Nano 2 is NVIDIA’s ultra-compact LLM tuned for on-device and edge deployment. It delivers fast instruction following, coding help, and reasoning with low memory use, supports tool/function calling and structured (JSON) outputs, and runs efficiently on Jetson, RTX laptops/desktops, and server GPUs.

  💻Coding🔍Code reviews💡Coding help🐍Python lessons
  Text

  Released 9mo ago
• Gen 7
  Kimi K2 Instruct

  By Moonshot AI

  Kimi-K2-Instruct-0905 is Moonshot AI’s upgraded Kimi K2 instruct model, a 1T-parameter MoE system with 32B active parameters. It is tuned for agentic coding, better real-world coding performance, and stronger frontend generation without the long-thinking behavior of the dedicated thinking variant

  💬Chatting🔍Code reviews🔍Advanced reasoning🖥️Frontend development💻Conversational coding
  Text

  Released 11mo ago
• Gen 3
  moondream 2

  By Moondream

  Moondream 2 is a small open vision-language model designed to run efficiently across many environments. It is the previous-generation Moondream model, released under Apache 2.0, and is positioned as a lightweight image-text model for practical multimodal use where smaller size and deployability matter more than maximum frontier scale.

  🖼️Image generation🔍Code reviews🔍Image recognition
  Multimodal

  Released 11mo ago
• Gen 7 Qwen
  Qwen 3

  By Alibaba

  Qwen 3 is Alibaba’s flagship general model for strong reasoning, coding, and multilingual use. It supports long context, tool and function calling, streaming, and clean JSON output for RAG and agents.

  💻Coding🔍Code reviews
  Text

  Released 1y ago
• Gen 7 Nemotron
  Llama 3.1 Nemotron Ultra

  By NVIDIA

  Llama 3.1 Nemotron Ultra is an NVIDIA-optimized deployment of Meta’s Llama 3.1, packaged for high-throughput production. It delivers strong reasoning and coding, long-context support (≈128K), tool/function calling, and JSON mode—served as a fast, scalable NIM for apps and agents.

  💻Coding🔍Code reviews🇯🇵Japanese lessons
  Text

  Released 1y ago
• Gen 7 Qwen
  Qwen 2.5 Plus

  By Alibaba

  Qwen 2.5 Plus is the balanced tier that trades a bit of speed for higher accuracy and stronger coding and reasoning.

  💻Coding✉️Email writing🔍Code reviews
  Text

  Released 1y ago
• Gen 7
  Devin 1.2

  By Cognition AI

  Devin 1.2 is a Devin update focused on better reasoning with repository context, plus new collaboration and admin features like Slack audio messages and enterprise account management.

  💻Software development guidance🔍Code reviews📊Project management
  Text

  Released 1y ago
• Gen 7
  DBRX Base

  By Databricks

  DBRX Base is Databricks’ open-weight Mixture-of-Experts foundation model (≈132B total parameters) built for long-context reasoning and coding. It’s the pretrained (unaligned) checkpoint—ideal for domain fine-tuning, RAG, and custom alignment—while DBRX Instruct is the chat-ready version.

  💻Coding🔍Code reviews🚀Web3 seo consultancy🇯🇵Japanese lessons
  Text

  Released 2y ago