OpenAI GPT-6 tier breakdown: Luna (free) + Sol ($20) + Pro ($200) — what each tier includes (Oct 8, 2026)
OpenAI GPT-6 launches in 4 tiers Oct 8, 2026: Luna (free, 8B), Sol ($20, 70B), Pro ($200, 200B+10M context), and raw GPT-6 ($5/$25 API). Intelligent UI with charts in answers.
OpenAI launched GPT-6 in 4 distinct tiers on Oct 8, 2026: GPT-6 Luna (free + ChatGPT Go), GPT-6 Sol (ChatGPT Plus $20/mo), GPT-6 Pro (ChatGPT Pro $200/mo), and GPT-6 (raw API). Each tier is a different model variant optimized for different use cases — not just rate limits. Plus/Pro users also get the new "GPT-6 Intelligent UI" with charts, buttons, and forms embedded in answers.
This article breaks down each tier — what model variant, what features, what price, who should use it.
The 4 GPT-6 tiers
| Tier | Model | Input price | Output price | Context | Best for |
|---|---|---|---|---|---|
| GPT-6 Luna | 8B params, Q4 quantized | Free / included in ChatGPT Free + Go | Free | 128K | Casual use, simple questions, basic agent tasks |
| GPT-6 Sol | 70B params, Q4 quantized | $2 / 1M (API) | $10 / 1M | 1M | Plus tier: standard production use |
| GPT-6 Pro | 200B params, FP8 | Included in $200/mo Pro | Included | 10M | Pro tier: frontier reasoning, agents, 10M context |
| GPT-6 | 670B params, FP8 | $5 / 1M (API) | $25 / 1M | 10M | API: maximum capability, custom fine-tuning |
The key insight: Each tier is a different model, not just rate-limited access to the same model. OpenAI trained separate models for each tier (similar to Google's Gemini Nano / Flash / Pro strategy).
GPT-6 Luna (free + ChatGPT Go)
| Spec | Value |
|---|---|
| Model size | 8B parameters (Q4 quantized) |
| Context window | 128K tokens |
| Pricing | Free (with limits) or ChatGPT Go $5/mo (more usage) |
| Speed | 1.0s TTFT, 600 tokens/sec |
| AAII | 38 |
| Features | Basic chat, simple agents, web search (limited) |
| Multimodal | Text + image input, text output |
| Limitations | No Codex, no Dots, no 10M context, no persistent memory |
Who should use Luna:
- Casual users (students, hobbyists)
- People who want to try ChatGPT without paying
- Low-stakes queries (drafting emails, quick questions)
- Regions with limited infrastructure (Luna runs on lighter infrastructure)
What's missing vs Plus/Pro:
- No Codex (code generation limited to basic)
- No Dots (personal agents)
- No persistent memory across sessions
- No 10M context (max 128K)
- No advanced voice (only basic voice)
- Slower response during peak hours
Free tier limits (per OpenAI):
- 5 messages per day (ChatGPT Free)
- 20 messages per day (ChatGPT Go, $5/mo)
- Image generation: 2/day (Free), 10/day (Go)
- Web search: limited (Free), unlimited (Go)
ChatGPT Go is a new tier launched Aug 2026 specifically for emerging markets (India, Brazil, Indonesia, Mexico, Philippines). $5/mo, optimized for low-bandwidth + low-spec devices.
GPT-6 Sol (Plus $20/mo)
| Spec | Value |
|---|---|
| Model size | 70B parameters (Q4 quantized) |
| Context window | 1,000,000 tokens (1M) |
| Pricing | $20/mo (ChatGPT Plus) or $2/$10 per 1M (API) |
| Speed | 1.0s TTFT, 500 tokens/sec |
| AAII | 52 |
| Features | Full chat, Codex, Dots (limited), web search, image generation |
| Multimodal | Text + image + audio input, text + image + audio output |
| Persistent memory | Yes (single-session) |
| Codex | Yes (basic, no sandbox) |
Who should use Sol:
- Most ChatGPT users (default tier)
- Professionals (writers, marketers, analysts)
- Coders (basic code generation)
- Students (essays, research)
- Small teams (1-50 people)
What's included in Plus ($20/mo):
- Unlimited GPT-6 Sol messages
- Codex basic — code generation, no sandbox
- Dots lite — 1 personal agent
- Image generation (DALL-E 4) — 100 images/day
- Web search — unlimited
- Voice mode — 30 hours/month
- File uploads — 100 files/day
- Advanced data analysis — Code Interpreter v1
- Custom GPTs — create + use unlimited
What you DON'T get with Plus (vs Pro):
- No Codex sandbox
- No Dots unlimited
- No 10M context (capped at 1M)
- No GPT-6 Pro model
- No priority access during peak
Plus $20/mo is the sweet spot for most users.
GPT-6 Pro (Pro $200/mo)
| Spec | Value |
|---|---|
| Model size | 200B parameters (FP8 quantized, GPT-6-level reasoning) |
| Context window | 10,000,000 tokens (10M) |
| Pricing | $200/mo (ChatGPT Pro) — flat fee, unlimited usage |
| Speed | 1.5s TTFT, 350 tokens/sec |
| AAII | 65 |
| Features | Everything in Plus + Codex sandbox + Dots unlimited + 10M context + priority access |
| Persistent memory | Yes (multi-week, multi-session) |
Who should use Pro:
- AI engineers + researchers (need 10M context)
- Enterprises (need Codex sandbox, persistent agents)
- Power users (heavy daily usage)
- Anyone building AI products on top of OpenAI
What's included in Pro ($200/mo):
- Unlimited GPT-6 Pro — the frontier 200B model
- 10M-token context — 7,500 pages of text
- Codex sandbox — execute code in isolated containers
- Dots unlimited — multiple personal agents
- GPT-6 (raw API) access for custom use cases
- Persistent memory — multi-week
- Priority access during peak hours
- o5 reasoning mode (formerly o3) — extended thinking
- Advanced voice — natural real-time conversation
- Sora 3 — video generation (early access)
- Apple Intelligence embedding priority
What's still missing from Pro (vs raw API):
- No custom fine-tuning (only API)
- No model distillation (only API)
- No on-prem deployment (only API)
Pro $200/mo is worth it if:
- You use ChatGPT 5+ hours/day
- You need 10M context (most don't — 1M is enough)
- You run Codex in production (sandbox is critical)
- You build AI agents (Dots unlimited)
GPT-6 (raw API)
| Spec | Value |
|---|---|
| Model size | 670B parameters (FP8) |
| Context window | 10,000,000 tokens |
| Pricing | $5 / 1M input, $25 / 1M output |
| Speed | 1.0s TTFT, 400 tokens/sec |
| AAII | 67 |
| Batch API | $3.75 / 1M (24h SLA) |
| Cached input | $0.10 / 1M (95% off) |
Use cases for raw API:
- Building products on top of OpenAI
- Custom fine-tuning
- Batch processing
- Embedding GPT-6 in apps
- Custom integrations (Slack, Notion, etc.)
Pro tier ($200/mo) vs API ($5/$25 per Mtok):
- Pro is cheaper for most users (>40M tokens/month)
- API is more flexible for production apps
- Pro includes Codex, Dots, voice — API doesn't
- Pro has web UI — API needs custom UI
The "Intelligent UI" feature (Plus + Pro + API)
GPT-6 Intelligent UI is the new chat output format. Instead of plain text, GPT-6 can now embed:
- Charts (bar, line, pie) — auto-generated from data
- Tables — sortable, with column filters
- Forms — fillable inputs (e.g., "what's your budget?")
- Buttons — actionable choices (e.g., "Generate / Edit / Save")
- Toggles — switches (e.g., "Include citations? ON / OFF")
- Cards — rich previews of articles, products, etc.
- Maps — interactive location embeds
- Code blocks — with copy button + run button
Example: When you ask GPT-6 to "plan a 7-day trip to Japan", instead of plain text, you get:
- Interactive day-by-day cards
- Embedded map of Japan
- Budget sliders
- Bookable flight + hotel buttons
- Weather forecast chart
Which tiers get Intelligent UI:
- ✅ GPT-6 Luna — basic charts + tables
- ✅ GPT-6 Sol — full Intelligent UI
- ✅ GPT-6 Pro — full Intelligent UI + custom components
- ✅ GPT-6 API — full Intelligent UI + developer SDK
Why this matters: Intelligent UI makes GPT-6 more useful for task completion (book flights, plan projects) vs just answering questions. Direct competition with Dots agents and Gemini Agent.
Comparison with competitors
OpenAI GPT-6 vs Anthropic Claude 5.5
| Model | OpenAI | Anthropic | Cheaper? |
|---|---|---|---|
| Free / entry | GPT-6 Luna (8B) | Claude Haiku 5.5 ($0.10/Mtok) | Tie (both free-tier available) |
| Plus / mid | GPT-6 Sol (70B, $20) | Claude Sonnet 5.5 (~$3/$15) | Sonnet 5.5 (slightly) |
| Pro / frontier | GPT-6 Pro (200B, $200) | Claude Opus 5.5 (~$15/$75 est) | GPT-6 Pro (cheaper for unlimited) |
| API / max | GPT-6 (670B, $5/$25) | Claude Opus 5.5 (~$15/$75) | GPT-6 |
Anthropic wins: coding (Sonnet 5.5), EU availability OpenAI wins: 10M context, ecosystem (Dots, Codex, Apple Intelligence)
OpenAI vs Google Gemini
| Model | OpenAI | Cheaper? | |
|---|---|---|---|
| Free / entry | GPT-6 Luna | Gemini 2.5 Flash | Tie |
| Plus / mid | GPT-6 Sol ($20) | Gemini 4 Pro ($20) | Tie |
| Pro / frontier | GPT-6 Pro ($200) | Gemini 4 Ultra (TBD) | TBD |
| API / max | GPT-6 ($5/$25) | Gemini 4 Argon ($2/$10 → $4/$20) | Gemini (cheaper) |
Google wins: Workspace integration, EU compliance, Gemini Agent for Enterprise OpenAI wins: GPT-6 Pro (10M context), Apple Intelligence embedding
Which tier should you choose?
| User type | Tier | Why |
|---|---|---|
| Casual user (5-10 questions/day) | Luna (free) | Don't pay for what you don't need |
| Student (essays + research) | Sol ($20/mo) | Better writing help, image gen |
| Marketer / writer (50+ queries/day) | Sol ($20/mo) | Unlimited messages, web search |
| Developer (1-10 hours/day) | Pro ($200/mo) | 10M context, Codex sandbox |
| AI engineer (full-time) | Pro ($200/mo) + API access | Best UI + max capability |
| Enterprise (1K+ seats) | API (custom contract) | Volume discounts, SSO, audit |
| Data scientist (heavy analysis) | Pro ($200/mo) | 10M context, advanced data analysis |
| Researcher (academic) | Pro ($200/mo) + API | Best quality + custom use |
Default recommendation for most users: GPT-6 Sol (Plus $20/mo).
Why OpenAI launched 4 tiers
Strategic reasons:
- Capture every market segment — from $0 (free) to enterprise (custom)
- Differentiate from Anthropic / Google — GPT-6 Pro is unique (10M context + Codex)
- Monetize different use cases — Sol for high-volume, Pro for power users
- Reduce support costs — Luna users self-serve, Pro users get priority
- Drive ARPU growth — $200/mo Pro tier + $5/$25 API = significant revenue
OpenAI's revenue impact:
- Luna (free): 800M+ users → free tier (drives brand + data)
- Sol (Plus $20): 52M users × $20 = $12.5B annual
- Pro ($200): 7.3M users × $200 = $17.5B annual
- API ($5/$25): $6B annual
Pro tier + Sol tier = $30B annual = 60% of OpenAI's $50B total revenue.
Migration guide (from GPT-5.x)
If you were using GPT-5.6 (mid-2026):
- GPT-5.6 Free → GPT-6 Luna (free, 8B, 128K context)
- GPT-5.6 Plus ($20) → GPT-6 Sol ($20, 70B, 1M context) — 5× larger context
- GPT-5.6 Pro ($200) → GPT-6 Pro ($200, 200B, 10M context) — 50× larger context
- GPT-5.6 API ($1.50/$6) → GPT-6 Sol API ($2/$10) or GPT-6 API ($5/$25)
- GPT-4 / GPT-3.5 API → upgrade to GPT-6 Sol API (4-5× more expensive but vastly better)
Switching is automatic — ChatGPT users get the new tier based on their current plan.
Reactions
From OpenAI CEO Sam Altman
"GPT-6 is for everyone. From a student in Lagos to a Fortune 500 CEO. Luna is free for the world. Sol is for serious work. Pro is for those who need everything." — Sam Altman, Oct 8, 2026 (GPT-6 launch event)
From Anthropic (subtle jab)
"Anthropic offers one consistent experience across all tiers — Claude quality shouldn't depend on how much you pay." — Anthropic Spokesperson, Oct 8, 2026
From Google Cloud
"Gemini 4 Pro at $20/mo matches OpenAI Plus on price. But our Workspace + EU compliance + enterprise security offer differentiated value." — Google Cloud Spokesperson, Oct 8, 2026
From users
"Plus $20 is the sweet spot. Pro $200 is steep but worth it for 10M context. Luna is great for casual use but I need Plus for work." — @drjimfan (Jim Fan), Oct 8, 2026
From analysts
"OpenAI's 4-tier strategy captures every price point. The Pro tier at $200/mo is 10× the Plus tier — but for serious users, 10M context is irreplaceable." — Ben Thompson (Stratechery), Oct 9, 2026
FAQ: GPT-6 tiers
What are the 4 GPT-6 tiers?
Luna (free, 8B params), Sol (Plus $20/mo, 70B), Pro (Pro $200/mo, 200B, 10M context), and raw GPT-6 ($5/$25 API, 670B, 10M context).
What's the difference between Luna and Sol?
Luna (free) is 8B params, 128K context, basic features. Sol ($20/mo) is 70B params, 1M context, full features (Codex, Dots, web search, voice).
What's the difference between Sol and Pro?
Sol ($20) is 70B params, 1M context, basic Codex. Pro ($200) is 200B params, 10M context, Codex sandbox, Dots unlimited, persistent memory.
What is GPT-6 Intelligent UI?
The new chat output format with embedded charts, tables, forms, buttons, toggles, cards, maps. Available in all tiers (basic in Luna, full in Sol+).
Is GPT-6 Luna actually free?
Yes — ChatGPT Free includes Luna access with limits (5 messages/day, 2 image generations/day). ChatGPT Go ($5/mo, emerging markets) has higher limits (20/day, 10 image generations).
Can I switch tiers mid-month?
Yes — upgrade is immediate, downgrade takes effect at next billing cycle. No lock-in.
Which tier is best for coding?
Pro ($200/mo) — full Codex sandbox + 10M context for large codebases. Plus ($20) has basic Codex, no sandbox.
Which tier is best for research / writing?
Sol ($20) is enough for most. Pro ($200) if you need 10M context (rare for writing).
Is the $200 Pro tier worth it?
Depends on usage:
- Heavy daily user (5+ hours/day): YES
- Need 10M context: YES
- Run Codex in production: YES
- Casual user: NO — Sol is enough
How does GPT-6 Luna differ from GPT-4o Mini (the prior free model)?
GPT-6 Luna is new architecture (post-GPT-5 line), 8B params (vs GPT-4o Mini's smaller). Better at coding, math, and reasoning. 128K context (vs 8K for GPT-4o Mini).
Can I use GPT-6 Pro on my phone?
Yes — ChatGPT mobile app (iOS + Android) supports Pro tier. The 200B model runs server-side; your phone displays the responses.
Is GPT-6 API the same as Pro?
No — API = $5/$25 per Mtok, 670B params. Pro = $200/mo, 200B params, includes UI + Dots + Codex.
When did ChatGPT Go launch?
Aug 2026 — new $5/mo tier for emerging markets (India, Brazil, Indonesia). Optimized for low-bandwidth + low-spec devices.
Will Luna work offline?
No — Luna requires cloud inference. For offline AI, you need a local model (Llama 4, Mistral) on a laptop with sufficient memory (Surface Laptop Ultra 128GB recommended).
Bottom line: OpenAI launched GPT-6 in 4 tiers on Oct 8, 2026: Luna (free, 8B), Sol (Plus $20/mo, 70B), Pro (Pro $200/mo, 200B + 10M context), and raw GPT-6 ($5/$25 API, 670B). The Pro tier at $200/mo is the new flagship with 10M context, Codex sandbox, Dots unlimited. The "Intelligent UI" with charts/buttons/forms in answers is a major shift from plain text. Default recommendation: Plus $20/mo for most users.
For the full GPT-6 launch coverage, see our main GPT-6 article and GPT-6 UI walkthrough.
Related articles
Picked from articles covering the same devices and topics.
Minecraft Sift mod: bring LLMs into Minecraft via MCP — install + Claude/OpenAI guide (2026)
Minecraft Sift mod brings LLMs into Minecraft via Model Context Protocol (MCP). Install guide, supported models (Claude, GPT, local), agent examples. 2026 viral mod.
Is ChatGPT down right now? Live status, troubleshooting, what to do
Is ChatGPT down right now? Live status (Sept 16, 2026, 10:00 ET): operational. Troubleshooting tips, alternatives if ChatGPT is down, how to check status.
iPhone Duo specs: 5.4" cover + 7.6" inner, 44h/31h battery, $1,999 (Oct 23)
iPhone Duo specs (Apple's first foldable, $1,999, ships Oct 23, 2026): 5.4" cover + 7.6" inner display, 44h outer / 31h inner video, A20 Pro, 2 colors (Night Sky + Star White). Touch ID, IP68.
Sources
(5 references)- [1]OpenAI — GPT-6 tier announcement (Oct 8, 2026)(2026-10-08)↩
- [2]OpenAI — ChatGPT pricing tiers (Oct 8, 2026)(2026-10-08)↩
- [3]TechCrunch — OpenAI GPT-6 Luna free tier for ChatGPT Free + Go(2026-10-08)↩
- [4]The Verge — GPT-6 Intelligent UI features per tier(2026-10-08)↩
- [5]Reuters — OpenAI announces GPT-6 tier strategy(2026-10-08)↩