Frontier models, one private account
35 models from Google, OpenAI, Alibaba, Moonshot AI, Paper Instruments, DeepSeek, Meta, xAI, Z.ai, Perplexity - an optional paid service, billed per use, with the price of every conversation shown as you go. You sign in with your Flowsta Identity: an identity you create and keep yourself, in your own Flowsta Vault with 24 words as the key - the most private way to use frontier models. No provider accounts, and everything local stays local.
Gemini 4 Neon
Google · Nov 2026Google's newest Gemini 4 model, with a huge context. Reads images as well as text.
GPT-6.2 Astra
OpenAI · Nov 2026OpenAI's newest flagship - the strongest model for hard problems, coding, and research, at a fifth less than GPT-6 Astra.
36 of 36 right on our questions · 3 s an answer · 15 cents for all 36
GPT-6.2 Sol
OpenAI · Nov 2026OpenAI's flagship for most work - close to GPT-6.2 Astra on coding and hard problems, at a quarter of Astra's price.
36 of 36 right on our questions · 2 s an answer · 4 cents for all 36
Qwen4 Max
Alibaba · Nov 2026Alibaba's newest flagship - a strong all-rounder with vision and 1M context.
36 of 36 right on our questions · 4 s an answer · 5 cents for all 36
Kimi K3.1 Pro
Moonshot AI · Oct 2026Moonshot's newest flagship - always thinks things through, huge context, strong with images.
36 of 36 right on our questions · 7 s an answer · 15 cents for all 36
Kimi K3.1 Pro Highspeed
Moonshot AI · Oct 2026Kimi K3.1 Pro at two to three times the speed.
36 of 36 right on our questions · 4 s an answer · 28 cents for all 36
Ultramarine 1 Flash
Paper Instruments · Oct 2026A model tuned for finance, law and consulting work, served from US data centers by Fireworks. Chat only.
36 of 36 right on our questions · 2 s an answer · 1 cents for all 36
DeepSeek V4.1 Flash
DeepSeek · Oct 2026The fast, ultra-cheap lane - quick answers for everyday questions, now with images.
36 of 36 right on our questions · 2 s an answer · 1 cents for all 36
DeepSeek V4.1 Flash (US)
DeepSeek · Oct 2026DeepSeek's fast, low-cost model, served from US data centers by Fireworks - quick everyday answers, with images.
36 of 36 right on our questions · 1 s an answer · 2 cents for all 36
Gemini 4 Argon
Google · Oct 2026Google's frontier model - complex coding, agents, math and science, with a huge context.
36 of 36 right on our questions · 18 s an answer · 16 cents for all 36
GPT-6.1 Terra
OpenAI · Oct 2026OpenAI's fast, efficient middle model - strong at visual reasoning and clear writing, at half GPT-6.1 Sol's price.
36 of 36 right on our questions · 2 s an answer · 2 cents for all 36
Muse Wonder 1.0
Meta · Oct 2026Meta's frontier agent model - long tasks, tools in parallel, and documents, pictures and video as input.
36 of 36 right on our questions · 4 s an answer · 44 cents for all 36
Qwen3.8 Omni Flash
Alibaba · Oct 2026Alibaba's very low-cost model - quick answers, good with images.
36 of 36 right on our questions · 5 s an answer · under 1 cent for all 36
GPT-6 Luna
OpenAI · Sep 2026OpenAI's fast, low-cost tier - great for everyday questions, at half the price of GPT-5.6 Luna.
36 of 36 right on our questions · 2 s an answer · under 1 cent for all 36
Grok 4.7
xAI · Sep 2026xAI's newest frontier model - top-tier coding, agent work, and chat, at the same price as before.
36 of 36 right on our questions · 10 s an answer · 31 cents for all 36
Grok 4.7 (Web)
xAI · Sep 2026xAI's newest frontier model with live web search - current info, cited sources.
36 of 36 right on our questions · 10 s an answer · 31 cents for all 36
Muse Spark Open
Meta · Sep 2026Meta's open-weights model for coding and agents, served from US data centers by Fireworks.
36 of 36 right on our questions · 4 s an answer · 20 cents for all 36
GPT-6 Astra
OpenAI · Sep 2026OpenAI's earlier flagship for hard problems, coding, and research. GPT-6.2 Astra follows it at a fifth less.
36 of 36 right on our questions · 3 s an answer · 25 cents for all 36
Muse Spark 1.3
Meta · Sep 2026Meta's frontier model - long tasks, coding, and agents at a fraction of flagship prices.
36 of 36 right on our questions · 9 s an answer · 18 cents for all 36
Gemini 3.8 Flash
Google · Sep 2026Google's fast, low-cost model - quick everyday answers, strong with images and long documents.
36 of 36 right on our questions · 4 s an answer · 5 cents for all 36
GLM-5.3 Flash (US)
Z.ai · Aug 2026Z.ai's fast, low-cost model with images, served from US data centers by Fireworks.
36 of 36 right on our questions · 1 s an answer · 1 cents for all 36
GLM-5.3
Z.ai · Aug 2026Z.ai's newest flagship - a step up in coding and agent work, 1M context.
36 of 36 right on our questions · 4 s an answer · 7 cents for all 36
GLM-5.3 (US)
Z.ai · Aug 2026Z.ai's flagship, served from US data centers by Fireworks - strong at coding and long agent work, 1M context.
36 of 36 right on our questions · 2 s an answer · 10 cents for all 36
GLM-5.3 Flash
Z.ai · Aug 2026Z.ai's fast, low-cost model with images, 1M context.
35 of 36 right on our questions · 6 s an answer · under 1 cent for all 36
GLM-5.3 FlashX
Z.ai · Aug 2026Z.ai's FlashX tier of GLM-5.3 - low cost, with images and 1M context.
36 of 36 right on our questions · 3 s an answer · 2 cents for all 36
DeepSeek V4
DeepSeek · Aug 2026Frontier-level reasoning and coding at a low price.
36 of 36 right on our questions · 4 s an answer · 6 cents for all 36
Kimi K3
Moonshot AI · Jul 2026Moonshot's flagship - deep reasoning and top-tier coding, huge context, strong with images.
36 of 36 right on our questions · 7 s an answer · 24 cents for all 36
Kimi K3 (US)
Moonshot AI · Jul 2026Moonshot's flagship, served from US data centers by Fireworks - deep reasoning, top-tier coding, huge context.
36 of 36 right on our questions · 3 s an answer · 27 cents for all 36
GPT-5.6 (Web)
OpenAI · Jul 2026GPT-5.6 with live web search - up-to-date, cited answers.
35 of 36 right on our questions · 2 s an answer · 6 cents for all 36
Kimi K2.7 Code
Moonshot AI · Jun 2026Moonshot's dedicated coding model - long-context programming, low cost.
36 of 36 right on our questions · 4 s an answer · 4 cents for all 36
Grok Build 0.1
xAI · May 2026xAI's fast, low-cost coding agent model.
36 of 36 right on our questions · 5 s an answer · 10 cents for all 36
Qwen3.7 Plus
Alibaba · May 2026Alibaba's balanced Qwen - strong everyday quality at a low price, with images.
36 of 36 right on our questions · 11 s an answer · 5 cents for all 36
Qwen3.6 Flash
Alibaba · Apr 2026Alibaba's fastest, cheapest Qwen - quick answers, with images.
36 of 36 right on our questions · 13 s an answer · 11 cents for all 36
Perplexity Sonar Pro
Perplexity · Mar 2025Deeper web research for harder questions.
35 of 36 right on our questions · 2 s an answer · 34 cents for all 36
Perplexity Sonar
Perplexity · Jan 2025Web-grounded answers with citations.
36 of 36 right on our questions · 2 s an answer · 24 cents for all 36
"Our questions" are the same set for every model, each with one answer that can be checked. The time and the cost are for answering all of them at our price, with the model's thinking counted: that is what separates models whose answers are equally right.
Prices are per million tokens, straight from the live catalog the app bills against. Nothing goes online without your consent: an AI set to Online and Offline uses these models, one set to Offline Only never does, and attachments ask first.
Prefer models that never leave your machine? Browse the offline catalog - 50 downloads, fully private.