Compare parts for your AI setup

Best for: Summaries, extraction, rewriting, support tasks, and short Q&A.

Not ideal for: Complex reasoning or long-context synthesis.

Published provider ceiling, not guaranteed visible answer length; reasoning, input, endpoint, or account conditions may reduce it.

Best for: Complex reasoning, proposal work, long-form synthesis, and high-stakes review.

Not ideal for: Bulk extraction, simple summaries, or cheap high-volume routing.

Published provider ceiling, not guaranteed visible answer length; reasoning, input, endpoint, or account conditions may reduce it.

Best for: Strong writing, coding, document analysis, and research drafts.

Not ideal for: High-volume, cost-sensitive extraction or classification.

Best for: Everyday reasoning, writing, code, and research summaries.

Not ideal for: Ultra-low-cost bulk extraction at scale.

Published provider ceiling, not guaranteed visible answer length; reasoning, input, endpoint, or account conditions may reduce it.

Best for: Long-context analysis, multimodal files, research, and document-heavy work.

Not ideal for: Short, simple chat where its context window and cost are overkill.

Published provider ceiling, not guaranteed visible answer length; reasoning, input, endpoint, or account conditions may reduce it.

Best for: Coding, reasoning, data analysis, and agent workflows.

Not ideal for: Cheap bulk tasks better handled by a budget model.

Published provider ceiling, not guaranteed visible answer length; reasoning, input, endpoint, or account conditions may reduce it.

Best for: Hard reasoning, complex coding, research synthesis, and tool-heavy workflows.

Not ideal for: Simple summaries or bulk extraction where a budget model is sufficient.

Published provider ceiling, not guaranteed visible answer length; reasoning, input, endpoint, or account conditions may reduce it.

Best for: Math, logic, coding help, and structured planning at lower cost.

Not ideal for: Long-form writing or multimodal work.

Best for: Code completion, repo tasks, and coding assistants.

Not ideal for: General-purpose chat.

Best for: Stronger self-hosted reasoning and coding than small distills.

Not ideal for: Low-memory local setups.

Published provider ceiling, not guaranteed visible answer length; reasoning, input, endpoint, or account conditions may reduce it.

Best for: Fast responses, multimodal summaries, and lightweight extraction.

Not ideal for: High-stakes reasoning or long-form synthesis.

Best for: Vision-language, screenshots, and lightweight visual reasoning.

Not ideal for: Text-only bulk work.

Best for: General chat, coding help, and structured outputs.

Not ideal for: Premium reasoning or multimodal analysis.

Best for: Conversational tasks, quick answers, and topical summaries when connected to search.

Not ideal for: Deep multi-step reasoning or heavy coding.

Best for: Lower-cost reasoning, planning, code review, and long-form answers.

Not ideal for: Multimodal tasks or latency-critical chat.

Best for: Chat, writing, summaries, and simple reasoning.

Not ideal for: Hard reasoning, complex coding, or tool-heavy agents.

Best for: Chat, writing, summaries, and simple reasoning.

Not ideal for: Hard reasoning, complex coding, or tool-heavy agents.

Best for: Coding, extraction, summarization, and cost-sensitive reasoning.

Not ideal for: Multimodal work or polished long-form writing.

Best for: Local/simple assistant, classification, extraction, and summaries.

Not ideal for: Hard reasoning, coding agents, or long-context work.

Reported download size is storage evidence, not a RAM or VRAM requirement. The separate 16 GB · LM Studio minimum is source-backed, but still does not prove complete hardware fit.

Best for: Local/simple assistant, classification, extraction, and summaries.

Not ideal for: Hard reasoning, coding agents, or long-context work.

Best for: Local multimodal tasks with lower active-parameter cost.

Not ideal for: Tasks needing maximum local quality.

Best for: Self-hosted general assistant, RAG, and internal tools.

Not ideal for: Frontier reasoning or multimodal tasks.

Reported download size is storage evidence, not a RAM or VRAM requirement. The separate 40 GB · LM Studio minimum is source-backed, but still does not prove complete hardware fit.

Best for: Self-hosted general assistant, RAG, and internal tools.

Not ideal for: Frontier reasoning or multimodal tasks.

Best for: Self-hosting, customization, and privacy-aware workloads.

Not ideal for: Managed convenience or top-tier reasoning out of the box.

Best for: Agentic coding, tool-heavy workflows, and long-context tasks.

Not ideal for: Multimodal or vision work.

Best for: Better local chat and summaries than tiny models, at manageable cost.

Not ideal for: Frontier reasoning or coding.

Best for: Fast local chat, summaries, extraction, and light multimodal tasks.

Not ideal for: Complex reasoning or hard coding.

Best for: Fast chat, extraction, support automation, and simple code.

Not ideal for: Complex reasoning or research synthesis.

Reported download size is storage evidence, not a RAM or VRAM requirement. The separate 14 GB · LM Studio minimum is source-backed, but still does not prove complete hardware fit.

Best for: Fast chat, extraction, support automation, and simple code.

Not ideal for: Complex reasoning or research synthesis.

Best for: Enterprise assistants, synthetic data, reasoning, and internal RAG.

Not ideal for: Latency-critical chat or multimodal analysis.

Best for: Long-horizon coding, repo work, tool recovery, and local dev agents.

Not ideal for: General chat or lightweight summarization.

Best for: Local image/document understanding and multimodal assistants.

Not ideal for: Low-memory laptops.

Best for: Local visual perception, image/question tasks, and OCR-style workflows.

Not ideal for: Frontier multimodal reasoning.

Best for: Coding help, local agents, multimodal notes, and tool-heavy workflows.

Not ideal for: Tiny/on-device setups.

Reported download size is storage evidence, not a RAM or VRAM requirement. The separate 20 GB · LM Studio minimum is source-backed, but still does not prove complete hardware fit.

Best for: Coding help, local agents, multimodal notes, and tool-heavy workflows.

Not ideal for: Tiny/on-device setups.

Reported download size is storage evidence, not a RAM or VRAM requirement. The separate 20 GB · LM Studio minimum is source-backed, but still does not prove complete hardware fit.

Best for: Coding help, local agents, multimodal notes, and tool-heavy workflows.

Not ideal for: Tiny/on-device setups.

Reported download size is storage evidence, not a RAM or VRAM requirement. The separate 20 GB · LM Studio minimum is source-backed, but still does not prove complete hardware fit.

Best for: Coding help, local agents, multimodal notes, and tool-heavy workflows.

Not ideal for: Tiny/on-device setups.

Reported download size is storage evidence, not a RAM or VRAM requirement. The separate 20 GB · LM Studio minimum is source-backed, but still does not prove complete hardware fit.

Best for: Coding help, local agents, multimodal notes, and tool-heavy workflows.

Not ideal for: Tiny/on-device setups.

Reported download size is storage evidence, not a RAM or VRAM requirement. The separate 20 GB · LM Studio minimum is source-backed, but still does not prove complete hardware fit.

Best for: Coding help, local agents, multimodal notes, and tool-heavy workflows.

Not ideal for: Tiny/on-device setups.

Reported download size is storage evidence, not a RAM or VRAM requirement. The separate 20 GB · LM Studio minimum is source-backed, but still does not prove complete hardware fit.

Best for: Coding help, local agents, multimodal notes, and tool-heavy workflows.

Not ideal for: Tiny/on-device setups.

Best for: High-volume summaries, extraction, and AWS-native workloads.

Not ideal for: Complex reasoning or agentic coding.

Best for: Routing, classification, short extraction, and simple Q&A.

Not ideal for: Reasoning-heavy or long-form tasks.

Best for: Self-hosting, privacy-sensitive workflows, and customization.

Not ideal for: Managed convenience or multimodal work.

Best for: Private local assistant, reasoning with tool support, and low-cost hosted routes.

Not ideal for: Multimodal work.

Best for: Local enterprise assistant, instruction following, and JSON/tool tasks.

Not ideal for: Hard reasoning or multimodal work.

Best for: Low-latency local reasoning, chat, and tool workflows.

Not ideal for: Multimodal or vision tasks.

Best for: Small local reasoning, math, and short code help.

Not ideal for: Long-context or multimodal work.

Reported download size is storage evidence, not a RAM or VRAM requirement. The separate 8 GB · LM Studio minimum is source-backed, but still does not prove complete hardware fit.

Best for: Small local reasoning, math, and short code help.

Not ideal for: Long-context or multimodal work.

Best for: Local code help and software-agent experiments.

Not ideal for: Frontier coding benchmarks.

Best for: Bulk processing, extraction, classification, and simple coding.

Not ideal for: Deep reasoning or high-quality long-form writing.

Published provider ceiling, not guaranteed visible answer length; reasoning, input, endpoint, or account conditions may reduce it.

Best for: Bulk summarization, extraction, routing, and classification.

Not ideal for: Complex reasoning or coding.

Best for: Codebase exploration, editing workflows, and agentic coding.

Not ideal for: General chat or lightweight summarization.

Best for: Local/private assistants, multimodal summaries, and document/image understanding.

Not ideal for: Lowest-resource laptops or top-tier coding agents.

Best for: Local coding, reasoning, research drafts, and multimodal input.

Not ideal for: Very low-memory setups.