All LLMs

Open source and proprietary models, side by side.

OpenAIProprietary

GPT-4o

Multimodal flagship across text, vision, and audio.

Native multimodal I/OFunction callingVisionRealtime voice

AnthropicProprietary

Claude 3.5 Sonnet

Best-in-class coding and long-form reasoning.

ArtifactsComputer useTool callingVision

Google DeepMindProprietary

Gemini 1.5 Pro

Massive 2M-token context for whole-codebase reasoning.

2M token contextNative multimodalVideo understandingGrounding with Search

MetaOpen Source

Llama 3.1 405B

Frontier-class open weights model.

Llama Community License

Open weights128K contextTool useMultilingual

Mistral AIOpen Source

Mistral Large 2

European frontier model with strong code & math.

Mistral Research License

128K contextFunction callingJSON modeMultilingual

AlibabaOpen Source

Qwen 2.5

Family of permissive open models with strong code variants.

Apache 2.0 (most sizes)

Many sizesApache 2.0Code & math specialistsLong context

DeepSeekOpen Source

DeepSeek V3

Efficient MoE rivaling closed frontier models.

MIT (code) / Model License

Mixture of ExpertsFP8 trainingStrong reasoningOpen weights

MicrosoftOpen Source

Phi-3

Small models punching above their weight.

CompactMIT licensedStrong reasoning for sizeMobile-ready