We keep running into the same question: which model should I use for this? So we built the field guide.
Coding: Claude Opus / Sonnet, GPT-4.1, DeepSeek Coder, Qwen2.5-Coder, CodeLlama, StarCoder2, Phi-4, Devstral, gpt-oss.
General chat: GPT-4.1 mini, Claude Haiku 4, Gemini 2.5 Flash, Mistral Small 3, Qwen3-70B-Instruct, Llama-3.3-70B-Instruct.
Reasoning: o3, o4-mini, DeepSeek R1, Qwen3-Thinking, Claude Opus extended thinking, Gemini 2.5 Pro thinking.
Vision: Gemini 2.5 Pro, GPT-4.1, Claude Opus 4 with vision, Qwen2.5-VL, Llama 3.2 Vision, Phi-4-multimodal.
Audio & speech: Speech Studio stack (Whisper + our local pipeline), Gemini Live, GPT-4o-realtime, Sesame CSM for voice cloning.
On-device: Phi-4, Gemma 3n, Qwen3-1.7B, SmolLM3, Apple's on-device models via Core ML.
Open weights you can run anywhere: Llama 3.3 70B, Mixtral 8x22B, Qwen3 235B-A22B, DeepSeek V3, Mistral Large 3, gpt-oss-120b, Kimi K2, Command R+.
Fine-tuning: Mistral, Together AI, Fireworks AI, Lamini, CloudMonkeys, Unsloth for local LoRA adapters.
The pattern: there is no 'best model'. There is the best model for your specific job, your latency budget, your data policy, and your wallet. This list exists so you can stop shopping and start shipping.