ChatGPT
GPT-6 Astra / 6.1 Sol / 6 Luna — OpenAI
OpenAI's reasoning powerhouse, now on GPT-6 and still split into three price tiers. Three tiers, and the middle one is the story: GPT-6.1 Sol lands one point behind the Astra flagship at a fifth of the price. The model that mainstreamed chain-of-thought for the rest of us — now with a 100x spread on the pricing sheet.
Released
Sep 2026
Context
1.05M tokens (Astra)
Starting Price
$20/mo (Plus)
Max Output
128K tokens (Astra)
01
Overview
ChatGPT needs no introduction — it's the model that put AI in everyone's hands. GPT-6 arrived across September 2026: Astra on September 3, Sol and Luna on September 22, then GPT-6.1 Sol on September 29. OpenAI kept the menu it introduced with GPT-5.6, but cut it to three dishes — there is no GPT-6 Terra.
Three tiers, and the middle one is the story. GPT-6 Astra is the flagship at $10 / $50 per million tokens, with a 1.05M token context window and 128K max output. GPT-6.1 Sol is what OpenAI calls “near-Astra performance for complex work at a lower cost” — $2 / $10, a fifth of Astra. GPT-6 Luna is the “most efficient model for focused, high-volume tasks” at $0.10 / $0.50. Astra to Luna is a 100x price spread.
The platform around the models moved too: the Agents API went into public beta on September 10, Codex picked up the GPT-6 models plus persistent Codex Cloud environments, and GPT-Image-2.5 took over image generation. On the way out: Sora's API shut down September 24, and custom GPTs retire December 11 in favor of plugins.
On Artificial Analysis's Intelligence Index v4.3.2, Astra scores 53 and GPT-6.1 Sol scores 52. One point, five times the price. Unless your evals show that point in your actual output, Sol is the default and Astra is the escalation.
Read the plan table carefully, because access is not what you'd assume. GPT-6 lives in ChatGPT's Work mode and in Codex on paid plans — not the regular Chat tab. Free and Go have run on GPT-5.6 Luna since August 6, with unlimited text chats. If you're on Free or Go, you're not using GPT-6.
02
Key Capabilities
Three Price Tiers, 100x Spread
Astra ($10/$50), GPT-6.1 Sol ($2/$10), Luna ($0.10/$0.50). Sol scores 52 to Astra's 53 on AA's Intelligence Index at a fifth the price — start on Sol (Luna for volume work) and move up only where evals justify it
Programmatic Tool Calling
Coordinates tools with JavaScript in a hosted runtime rather than round-tripping every call through the context window — fewer tokens, less latency on multi-step work
Native Computer Use
Control desktops directly — click, type, navigate software. End-to-end automation across tools until a task is finished.
1.05M Token Context
Astra ships a 1,050,000-token window with 128K max output. Requests past 272K input tokens bill the whole request at 2x input and 1.5x output
Agents API and Codex
The Agents API has been in public beta since September 10. Codex runs the GPT-6 models and picked up persistent Codex Cloud environments on September 29
Reasoning Tokens
Supports reasoning tokens for chain-of-thought workloads where deeper deliberation matters
Finance-Focused Enterprise Bundle
Native Excel and Google Sheets plugins for financial workflows
Persistent Multi-Round Search
More thorough web search across multiple sources
03
Pricing
| Tier | Price | Details |
|---|---|---|
| Free | $0 | Runs GPT-5.6 Luna (the default since Aug 6) with unlimited text chats. No GPT-6. |
| ChatGPT Go | $8/mo | Same model as Free: GPT-5.6 Luna, unlimited text chats. Still no GPT-6. |
| ChatGPT Plus | $20/mo | GPT-6 in Work mode and Codex. The real entry point for GPT-6. |
| ChatGPT Pro | $100 or $200/mo | Higher GPT-6 limits, and reportedly GPT-6 Astra Pro in ChatGPT. |
| ChatGPT Pro $500 | $500/mo | New at DevDay (Sep 29). 25x Plus usage — for people who live in Codex. |
| ChatGPT Business | $25/user/mo | GPT-6 in Work mode and Codex, reportedly including GPT-6 Astra Pro. |
| Enterprise | Custom | Full access, security, compliance features. |
| API — GPT-6 Astra | $10 / $50 per 1M tokens | Flagship. 1.05M context, 128K output. Pay for it only where Sol measurably falls short. |
| API — GPT-6.1 Sol | $2 / $10 per 1M tokens | Near-Astra quality (52 vs 53 on AA) at a fifth the price. The default for complex work. |
| API — GPT-6 Luna | $0.10 / $0.50 per 1M tokens | Cheapest. Focused, high-volume tasks — extraction, classification, transformations. |
| API — Long context | 2x input / 1.5x output | Past 272K input tokens, the multiplier applies to the whole request. |
| API — Batch / Flex | 50% off | For anything that can wait. Fast mode goes the other way at 2x. |
| API — GPT-5.6 (previous gen) | Sol $4 / $20 (promo) | Sol, Terra, and Luna stay on the API as the previous generation. Sol's promo rate runs at least through Nov 21. |
04
Best Use Cases
Desktop and browser automation via native computer use
Enterprise workflows involving spreadsheets, presentations, and documents
Complex research requiring persistent multi-round web searches
Professional coding in Codex — GPT-6.1 Sol gets near-Astra quality at a fifth the price
Tasks requiring long-context planning and execution
Multimodal document parsing and visual understanding
Organizations needing flexible pricing from free to enterprise
05
Pro Prompts
Strategic Planning
Plan a complete marketing strategy for a SaaS launch, including contingencies for each stage. Start with the riskiest assumptions.Computer Use
Open the Q4 sales spreadsheet, filter for deals over $50K, export a summary CSV, and draft an executive email highlighting the top 5 deals by revenue.Research Synthesis
Search for the latest developments in [field] from the past month. Cross-reference at least 5 sources. Highlight where sources disagree and rate your confidence in each finding.Code Review
Review this pull request for security, performance, and standards issues. Report every finding, including low-confidence and low-severity ones — don't filter for importance here. Tag each with a confidence level and severity so I can rank them myself.Tier Routing (GPT-6)
Here's my pipeline: [describe the steps]. For each step, tell me whether GPT-6 Astra, GPT-6.1 Sol, or GPT-6 Luna is the right tier and why. Treat Sol as the default, send high-volume, well-defined steps to Luna, and assume I'll only pay for Astra where the quality difference actually shows up in the output.06
Limitations
Hallucinations keep dropping generation over generation, but they aren't zero — and the remaining errors still land with confidence
Overactive safety guardrails can block legitimate professional requests (false positives)
Three tiers means three decisions. Defaulting everything to Astra is the easy mistake — and a 5x price mistake, given Sol lands one point behind it on the Intelligence Index
Long-context billing bites: past 272K input tokens, the whole request bills at 2x input and 1.5x output — the 1.05M window isn't uniformly priced
Astra at $10/$50 is premium pricing, and even Sol at $2/$10 gets undercut — Grok 4.7 ($2/$6) and Gemini 3.8 Flash ($0.75/$3.75 intro pricing through Dec 31) are cheaper for cost-sensitive workloads
GPT-6 is paid-only and lives in Work mode and Codex, not the regular Chat tab. Free and Go run GPT-5.6 Luna. Plus ($20/mo) is the real entry point for GPT-6
OpenAI plans to introduce ads in the free and Go tiers
Features are being taken away: Sora is gone (app closed Apr 26, API shut Sep 24), and custom GPTs retire Dec 11 in favor of plugins — if a workflow depends on a GPT, start migrating now
07
Multimodal Capabilities
Input
- Text
- Images (full-resolution support)
- File uploads (documents, PDFs, spreadsheets)
Computer Use
- Native desktop control
- Screenshots
- Mouse and keyboard input
Output
- GPT-Image-2.5 (Flare/Sunburst) image generation, since Sep 8
- Multi-round persistent web search
Not Supported
- No native audio or video input processing
- Separate Whisper/TTS models handle audio
08
Recent Updates
Sep 29, 2026
DevDay: GPT-6.1 Sol ships at $2 / $10, superseding GPT-6 Sol — near-Astra performance at a fifth the price. A new $500 Pro plan arrives with 25x Plus usage, and Codex Cloud gets persistent environments.
Sep 24, 2026
Sora's API shut down, five months after the app closed. If you built on it, that pipeline is done.
Sep 22, 2026
GPT-6 Sol and GPT-6 Luna released. Luna lands at $0.10 / $0.50 — and there's no Terra this generation.
Sep 8, 2026
GPT-Image-2.5 (Flare/Sunburst) launched as OpenAI's new image model
Sep 3, 2026
GPT-6 Astra released at $10 / $50 with 1.05M context and 128K output — Trusted Access first, then Plus, Pro, Business, Enterprise, and the API.
Aug 6, 2026
GPT-5.6 Luna replaced GPT-5.5 Instant as the Free and Go default, with unlimited text chats. (Correction: earlier versions of this page said Free and Go were still on 5.5 Instant after this date.)
Aug 21, 2026
GPT-5.6 Sol cut to $4 / $20 per million, down from $5 / $30 — promotional, now running at least through November 21. The Sol-to-Luna spread narrows from 25x to 20x, which makes Terra the more interesting default.
Jul 30, 2026
OpenAI adjusted GPT-5.6 pricing — Terra down 20%, Luna down 80%. Worth re-running your cost model if you priced it at launch.
Jul 9, 2026
GPT-5.6 became generally available — Sol, Terra, and Luna, all with 1.05M context and 128K max output
Jul 8, 2026
GPT-Live introduced — a new generation of voice models now powering ChatGPT Voice
May 5, 2026
GPT-5.5 Instant became the default ChatGPT model
Jul 2026
At launch, GPT-5.6 Sol led the Artificial Analysis Coding Agent Index at 80 points and scored 59 on Intelligence Index v4.1; SWE-bench Pro came in at 64.6% / 63.4% / 62.7% for Sol / Terra / Luna. Those v4.1 scores don't compare with today's v4.3.2 numbers.
Reported
The Wall Street Journal reports GPT-6.1 Astra was cancelled before release. Don't plan around an Astra refresh — GPT-6 Astra is the flagship you get.