Skip to content

Model Matrix

Your quick chooser for supported models in Fresh Focus AI. Canonical IDs (in code) are what you’ll use in settings, Tasks, and the AI tab. Display names match provider branding where possible.

Note

Specs below mirror your app’s configuration. Providers change limits over time; treat this page as operational guidance, not legal specs.


Quick picks

  • Default (balanced): gpt-5-mini — fast, reliable tool use, good price/perf.
  • Deep reasoning / STEM: gpt-o3, gpt-o3-mini, claude-opus-4-0, claude-3-7-sonnet-latest.
  • Ultra fast & cheap: gpt-5-nano, gemini-2.5-flash, gemini-2.5-flash-lite.
  • Multimodal (text+image input): gpt-4o, gpt-4o-mini, gpt-4.5-preview, claude-3.5-*, gemini-* (see tables).
  • Long context (≥ 200K): gpt-5, gpt-5-mini, gpt-o3, gpt-4.1*, claude-* (200K), gemini 2.x.
  • RAG & tools: gpt-5*, gpt-4o*, o-series, claude-*, gemini-2.5-pro, cohere_command-r(+).

Legend: Ctxt = context window (tokens), MaxOut = max output tokens, Mods = modalities, Tools = tool/function calling.


OpenAI

GPT-5 family

ID Name Ctxt MaxOut Mods Tools Notes
gpt-5 OpenAI GPT-5 400K 128K text Flagship reasoning/coding; strongest tool use; higher latency/cost.
gpt-5-mini OpenAI GPT-5 Mini 200K 32K text Recommended default — speed/cost balance, reliable JSON & tools.
gpt-5-nano OpenAI GPT-5 Nano 64K 8K text Ultra-low latency/cost; routing, tagging, simple agents.

O-series (reasoning)

ID Name Ctxt MaxOut Mods Tools Notes
gpt-o3 OpenAI o3 1M 100K text Top-tier deliberate reasoning; slow/expensive; STEM & proofs.
gpt-o3-mini OpenAI o3-mini 200K 100K text Faster/lighter o-series; great for code & math; text-only.
gpt-o1-mini OpenAI o1-mini 128K 65.5K text Fast reasoning under real-time constraints; coding-friendly.
gpt-o4-mini OpenAI o4-mini 1M 100K text Cost-efficient reasoning; step-by-step; may be slower on trivial asks.

GPT-4.x & Omni

ID Name Ctxt MaxOut Mods Tools Notes
gpt-4.5-preview OpenAI GPT-4.5 (Preview) 128K 16K text,image Large preview model; expensive; multimodal in; strong accuracy.
gpt-4 OpenAI GPT-4 8K 8K text,image Classic GPT-4; smaller context; higher cost vs 3.5.
gpt-4o OpenAI GPT-4o (Omni) 128K 16K text,image Multimodal Omni; strong multilingual; parallel tool calls.
gpt-4o-mini OpenAI GPT-4o Mini 128K 16K text,image Cheaper/faster Omni; good generalist multimodal.
gpt-4.1 OpenAI GPT-4.1 1M 32K text,image Very long context 4.x; parallel tools; Azure-first availability.
gpt-4.1-mini OpenAI GPT-4.1 mini 1M 32K text,image Mini 4.1; cost-efficient long-context.
gpt-4.1-nano OpenAI GPT-4.1 nano 1M 32K text,image Fastest 4.1 tier; small-task throughput.
gpt-4-turbo OpenAI GPT-4 Turbo 128K 4K text,image Cost-optimized GPT-4; improved accuracy; 4K output cap.
gpt-4-1106-preview OpenAI GPT-4 (1106 Preview) 128K 4K text,image Historical preview of Turbo features.

GPT-3.5 (legacy)

ID Name Ctxt MaxOut Mods Tools Notes
gpt-3.5-turbo OpenAI GPT-3.5 Turbo 4K 4K text Very fast & cheap; 2021 cutoff; basic tool use.
gpt-3.5-turbo-1106 OpenAI GPT-3.5 Turbo 1106 16K 16K text Extended context; JSON/parallel calls.
gpt-3.5-turbo-16k OpenAI GPT-3.5 Turbo 16K 16K 16K text Long-context variant; legacy.

Google (Gemini)

ID Name Ctxt MaxOut Mods Tools Notes
gemini-2.5-pro Gemini 2.5 Pro 1M 65.5K text,image,audio Highest-quality Gemini; strong reasoning & grounding.
gemini-2.5-flash Gemini 2.5 Flash 1M 32K text,image,audio Speed/cost-optimized; great for interactive UX.
gemini-2.5-flash-lite Gemini 2.5 Flash Lite 256K 8K text,image Lowest latency/cost; routing & summaries.
gemini-2.0-flash Gemini 2.0 Flash 1M 8K text,image,audio,video 2.0 workhorse; native tool use.
gemini-2.0-flash-lite Gemini 2.0 Flash Lite 1,048,576 8K text Fastest 2.0 tier; budget tasks.
gemini-1.5-pro-latest Gemini 1.5 Pro 8K text,image,audio,video Long-doc analysis; multimodal in; no tools.
gemini-1.5-flash-latest Gemini 1.5 Flash 8K text,image,audio,video High-throughput multimodal in; no tools.

Anthropic (Claude)

ID Name Ctxt MaxOut Mods Tools Notes
claude-opus-4-0 Claude 4.0 Opus 200K 32K text,image Most capable; premium; long context.
claude-sonnet-4-0 Claude 4.0 Sonnet 200K 64K text,image High-performance 4.0; strong vision.
claude-3-7-sonnet-latest Claude 3.7 Sonnet 200K text,image Hybrid reasoning; code execution loop.
claude-3-5-sonnet-latest Claude 3.5 Sonnet 200K text,image SOTA 2024; great writing/vision.
claude-3-5-haiku-latest Claude 3.5 Haiku 100K text,image Speed/cost tier; solid coding.
claude-3-opus-latest Claude 3 Opus 100K text,image Prior flagship; now superseded.
claude-3-sonnet-20240229 Claude 3 Sonnet (2024-02-29) 100K text Historical snapshot.
claude-3-haiku-20240307 Claude 3 Haiku (2024-03-07) 100K text Historical speed tier.
claude-2.1 Claude 2.1 200K 200K text Enterprise long-doc; reduced hallucinations.

xAI (Grok)

ID Name Ctxt MaxOut Mods Tools Notes
grok-4-0709 xAI Grok-4 256K text,image Flagship; strong math/reasoning; X integration.
grok-3-latest xAI Grok-3 131K text,image Enterprise-ready; coding/data extraction.
grok-3-mini-latest xAI Grok-3 Mini 131K text,image Lightweight “thinks before responding”.
grok-2-latest xAI Grok-2 131K text,image Legacy real-time model; X platform tie-ins.

Cohere

ID Name Ctxt MaxOut Mods Tools Notes
cohere_command Cohere Command 4K 4K text General instruction following; cost-effective.
cohere_command-r Cohere Command R 128K 4K text RAG/analytical Q&A; long context.
cohere_command-r-plus Cohere Command R+ 128K 4K text Higher quality RAG/Doc synthesis.
cohere_command-nightly Cohere Command (Nightly) 128K 4K text Experimental nightly build; unstable.

DeepSeek

ID Name Ctxt MaxOut Mods Tools Notes
deepseek-chat DeepSeek Chat 64K 8K text Fast MoE chat; coding support.
deepseek-reasoner DeepSeek Reasoner (R1) 64K 8K text Explicit chain-of-thought; slower, analytical.

Selection tips

  • Pick for the task: reasoning (o-series/Claude), fast iteration (mini/flash), multimodal (4o/Gemini), or bulk generation (nano/flash-lite).
  • Constrain outputs: ask for JSON schemas or markdown tables when exporting to Files, Drive, or Evernote.
  • Guide verbosity: use your app’s response controls (e.g., “reasoning effort,” “verbosity”) when available on GPT-5 family.
  • Budget: long contexts cost more; summarize first, then switch to a longer-context model only when needed.

Notes & caveats

  • Names vs IDs: We normalize display names (e.g., “GPT-4.1”) while keeping the canonical IDs in code.
  • Training data: Several entries note approximate cutoffs per provider descriptions in your config.
  • Deprecated: GPT-3.5 and some Claude 3.x snapshots are kept for compatibility; prefer newer tiers when quality matters.