List-price bands, speed, reasoning, and when-to-use guidance for chat and image models. Reference pricing only—not your Ethyx invoice.
Ethyx's recommended everyday chat picks—start here if you want our defaults.
Anthropic
Workload: Balanced
Claude Sonnet 5 is Ethyx’s balanced default—strong writing, analysis, and code without Opus cost. Use it for most projects; reserve Opus or Fable for maximum quality on hard problems.
List price: $$$
Speed: Standard
Supports extended reasoning
Sign up to start chatting with this model.
OpenAI
Workload: Light
GPT-5.6 Luna is the efficient GPT-5.6 tier for cost-sensitive, high-volume workloads. Step up to Terra or Sol when quality matters more than price.
List price: $$
Speed: Fast
Supports extended reasoning
Sign up to start chatting with this model.
xAI
Workload: Heavy
Grok 4.5 is xAI’s coding-focused model for agentic software, engineering, and workflow tasks. Use Grok 4.3 for general chat and mixed workloads.
List price: $$
Speed: Slower-paced
Supports extended reasoning
Sign up to start chatting with this model.
Sorted by access tier and workload. Perplexity Sonar models are included for web-grounded chat.
Anthropic
Workload: Very light
Haiku is Anthropic’s fast lane—summaries, short edits, and responsive everyday chat. Choose Sonnet when you need stronger writing or code review, or Opus/Fable for the hardest reasoning.
List price: $$
Speed: Instant
Non-reasoning focused
Sign up to start chatting with this model.
DeepSeek
Workload: Light
DeepSeek Chat offers affordable general conversation with solid quality for daily use. Choose DeepSeek Reasoner when problems need visible extended thinking.
List price: $
Speed: Fast
Non-reasoning focused
Sign up to start chatting with this model.
Gemini
Workload: Very light
Gemini 3.1 Flash-Lite remains available for high-volume chat where cost matters most. Prefer Gemini 3.5 Flash-Lite for the newer lite generation, or Gemini 3.6 Flash when you need stronger agentic and coding performance.
List price: $
Speed: Instant
Non-reasoning focused
Sign up to start chatting with this model.
Gemini
Workload: Very light
Gemini 3.5 Flash-Lite is Google’s current lite tier for high-volume chat and subagent work. Keep 3.1 Flash-Lite when you specifically need that generation; step up to Gemini 3.6 Flash for stronger agentic performance.
List price: $
Speed: Instant
Non-reasoning focused
Sign up to start chatting with this model.
OpenAI
Workload: Light
GPT-5.4 Mini balances capability and cost for coding helpers, subagents, and daily professional chat. It is a strong default when Nano feels too thin but GPT-5.4 full tier is more than you need.
List price: $
Speed: Fast
Non-reasoning focused
Sign up to start chatting with this model.
OpenAI
Workload: Very light
Reach for GPT-5.4 Nano when you need the lowest-cost OpenAI tier for quick drafts, classification, and simple Q&A. It trades depth for speed—step up to Mini, Luna, or GPT-5.4 when answers need more nuance.
List price: $
Speed: Instant
Non-reasoning focused
Sign up to start chatting with this model.
xAI
Workload: Very light
Use Grok 4.20 Non-Reasoning for quick Grok replies on straightforward prompts. It keeps cost and latency down—move to Grok 4.20 Reasoning, Grok 4.3, or Grok 4.5 when the task needs deeper analysis or coding.
List price: $$$
Speed: Instant
Non-reasoning focused
Sign up to start chatting with this model.
Perplexity
Workload: Light
Sonar in chat combines conversational replies with built-in web search and citations—great when answers should be grounded in current sources rather than model memory alone.
List price: $
Speed: Fast
Non-reasoning focused
Sign up to start chatting with this model.
Gemini
Workload: Light
Gemini 3.5 Flash remains a strong GA Flash model for agentic loops and coding. Prefer Gemini 3.6 Flash for the newest Flash generation unless you specifically need 3.5.
List price: $$
Speed: Fast
Supports extended reasoning
Sign up to start chatting with this model.
Gemini
Workload: Light
Gemini 3.6 Flash is Google’s latest Flash workhorse—strong for agentic loops, coding cycles, and responsive general chat with better token efficiency than 3.5 Flash.
List price: $$
Speed: Fast
Supports extended reasoning
Sign up to start chatting with this model.
xAI
Workload: Balanced
Grok 4.3 is xAI’s strong general model for mixed workloads that benefit from reasoning. Pair it with Sonar when you need citations, Grok 4.5 for coding-first work, or Grok 4.20 Non-Reasoning when speed dominates.
List price: $$
Speed: Standard
Supports extended reasoning
Sign up to start chatting with this model.
Kimi
Workload: Balanced
Kimi K2.6 shines on long-context work, visible thinking, coding, and multimodal chat. Choose it when threads are large or you want to follow the model’s reasoning steps.
List price: $$
Speed: Standard
Supports extended reasoning
Sign up to start chatting with this model.
DeepSeek
Workload: Heavy
DeepSeek Reasoner adds extended thinking for math, logic, and multi-step problems. Use DeepSeek Chat when you want faster, cheaper general conversation.
List price: $
Speed: Slower-paced
Supports extended reasoning
Sign up to start chatting with this model.
Gemini
Workload: Heavy
Gemini 3.1 Pro is Google’s top chat tier for difficult multimodal and analytical work. Gemini 3.6 Flash is the better fit when responsiveness and agentic speed matter more than depth.
List price: $$$
Speed: Standard
Supports extended reasoning
Sign up to start chatting with this model.
OpenAI
Workload: Heavy
GPT-5.3 Codex is OpenAI’s current agentic coding model for refactors, tool use, and long coding sessions. Prefer it over general chat models when the task is code-first; use Sonnet or Opus when prose quality matters more.
List price: $$$
Speed: Slower-paced
Supports extended reasoning
Sign up to start chatting with this model.
OpenAI
Workload: Heavy
GPT-5.4 remains a strong OpenAI frontier tier for demanding professional work. GPT-5.6 Terra/Sol are the newer family—use 5.5 or Sol when you need the absolute top OpenAI tier.
List price: $$$
Speed: Slower-paced
Supports extended reasoning
Sign up to start chatting with this model.
OpenAI
Workload: Heavy
GPT-5.6 Terra balances intelligence and cost for everyday professional work. Prefer Sol for the hardest OpenAI tasks, or Luna when volume and price dominate.
List price: $$$
Speed: Standard
Supports extended reasoning
Sign up to start chatting with this model.
xAI
Workload: Heavy
Grok 4.20 Reasoning is for analysis-heavy Grok workloads where non-reasoning Grok would feel too shallow. Consider Grok 4.3, Grok 4.5, or a frontier Claude/OpenAI model for the hardest tasks.
List price: $$$
Speed: Slower-paced
Supports extended reasoning
Sign up to start chatting with this model.
Perplexity
Workload: Heavy
Sonar Pro in chat delivers deeper Perplexity research with stronger synthesis and citations—use it when Sonar’s everyday web grounding is not enough.
List price: $$$
Speed: Standard
Supports extended reasoning
Sign up to start chatting with this model.
Anthropic
Workload: Very heavy
Claude Fable 5 is Anthropic’s most capable widely released model for demanding reasoning, long-horizon agentic coding, and high-autonomy work. Use Opus 5 when you want top Opus quality at lower cost than Fable.
List price: $$$$
Speed: Slower-paced
Supports extended reasoning
Sign up to start chatting with this model.
Anthropic
Workload: Very heavy
Claude Opus 4.8 remains available for complex reasoning, long documents, and premium writing. Prefer Claude Opus 5 for the newest Opus generation; use Sonnet for everyday work to save cost and time.
List price: $$$$
Speed: Slower-paced
Supports extended reasoning
Sign up to start chatting with this model.
Anthropic
Workload: Very heavy
Claude Opus 5 is Anthropic’s latest Opus tier for deep reasoning, agentic coding, and long-horizon work. Use Opus 4.8 when you specifically need that generation; prefer Sonnet for everyday work.
List price: $$$$
Speed: Slower-paced
Supports extended reasoning
Sign up to start chatting with this model.
OpenAI
Workload: Very heavy
GPT-5.5 remains OpenAI’s proven frontier model for the hardest tasks while GPT-5.6 Sol soaks. Prefer Sol when you want the newest flagship generation.
List price: $$$$
Speed: Slower-paced
Supports extended reasoning
Sign up to start chatting with this model.
OpenAI
Workload: Very heavy
GPT-5.6 Sol is OpenAI’s flagship GPT-5.6 model for complex reasoning and coding. Expect higher latency and cost than Terra; GPT-5.5 remains available while 5.6 soaks.
List price: $$$$
Speed: Slower-paced
Supports extended reasoning
Sign up to start chatting with this model.
Image generation models—use when you need visuals from a prompt rather than text chat.
xAI
Workload: Balanced
Grok Imagine is the fast xAI option for everyday image generation and edits—concepts, social visuals, and iterative tweaks. Choose GPT Image 2 when fidelity and fine detail matter more than speed.
OpenAI
Workload: Heavy
GPT Image 2 is OpenAI’s higher-quality image tier for detailed scenes, product mockups, and careful edits. Use Grok Imagine when you want quicker iteration at lower cost.
Row icons
List price
Speed
Reasoning
Privacy rating
No training per provider terms — under its commercial API terms, this provider does not train on your conversations.
No-retention requested — by default, Ethyx sends a per-request no-retention flag (store: false) with chat requests.
Caution — this provider may store and train on your messages. Avoid sharing sensitive details.
Privacy ratings describe provider policies, not a guarantee. Your real name and email are never sent raw—identity is pseudonymized or withheld—and by default we request no data retention where the provider API supports it.
