GPT-5.6 Sol Now Runs All of ChatGPT — Free Users Get Unlimited Chats

OpenAI just removed the choice most ChatGPT users didn't want to make. On August 6, 2026 (~5pm PT), @OpenAI announced it is folding ChatGPT's separate Instant and reasoning models into a single GPT-5.6 Sol experience for Plus and Pro users, while making text chat unlimited on GPT-5.6 Luna for Free and Go users starting the next day. The company's own framing: "We're making better intelligence easier to access in ChatGPT for everyone."

This lands a month after OpenAI's July 30 price cuts to Luna and Terra and five weeks after Sol's public GA rollout on July 9 — the same Sol/Terra/Luna tier system OpenAI introduced in June, now getting a consumer-facing simplification pass rather than a new generation number.

TL;DR — what actually changed

Is GPT-5.6 Sol replacing Instant and Thinking as separate models?

For Plus and Pro subscribers, yes. Until this update, ChatGPT routed everyday conversations between a fast Instant model and a slower reasoning model depending on the question — a split that dates back to earlier GPT-5.x generations. As of August 6, 2026, GPT-5.6 Sol powers both inside the everyday chat surface, replacing the two-model split with one consistent model and interaction pattern.

Practically, that means the response you get for "what's the capital of France" and the response you get for a multi-step tax question now come from the same underlying model, with Sol internally deciding — or the user deciding via the new slider — how much reasoning effort to apply. This mirrors how OpenAI already positioned Sol as the flagship of the Sol/Terra/Luna family launched in June: one capable model, tunable depth, rather than a hard cut between "fast and dumb" and "slow and smart."

Do free users really get unlimited chats now?

Yes, with one important scope limit: it's unlimited text chats, specifically. OpenAI's announcement states Free and ChatGPT Go plan users get unlimited text conversations with GPT-5.6 Luna, rolling out "starting tomorrow" relative to the August 6 announcement — meaning August 7, 2026. OpenAI did not extend the unlimited framing to voice, image generation, or other ChatGPT features in this announcement, so don't assume everything free became unlimited — just text chat.

This is a meaningful shift for a company that has spent much of 2026 tightening usage limits elsewhere — see explainx.ai's coverage of GPT-5.6 Sol's own usage-limit changes in July and the broader true cost of AI subscriptions across ChatGPT, Claude, and Gemini. Removing the text-chat cap for free users is a retention and top-of-funnel play more than a cost-cutting one — it makes the free tier materially more usable at the exact moment OpenAI is also making that tier's underlying model, Luna, meaningfully more accurate.

What does the 68% fewer factual errors claim actually measure?

OpenAI reports that in its own high-stakes factuality evaluation — covering finance, medicine, and law — the updated GPT-5.6 Sol produced 68% fewer responses with factual errors than GPT-5.5 Instant. Two things worth being precise about:

1. This is OpenAI's own benchmark, not an independently audited or third-party evaluation. As explainx.ai has noted when covering OpenAI's other vendor-published charts — like the Artificial Analysis Intelligence Index cited alongside July's price cuts — a company grading its own model's improvement is a real data point, but it isn't the same evidentiary weight as an independent audit. Treat 68% as a directional claim worth testing against your own workload, not a guaranteed real-world reduction.
2. The comparison baseline is GPT-5.5 Instant, an older-generation model — not a same-generation Sol variant. That's a meaningful comparison for anyone still on GPT-5.5-era ChatGPT, but it says less about how much better this specific Sol update is versus the Sol that already shipped in July.

For readers who want a general framework on reading vendor benchmark claims skeptically, explainx.ai's guide to reading AI benchmarks and complete AI benchmarks guide cover the pattern of self-reported factuality and capability claims across labs in 2026.

Where's the model picker — can I still choose old models?

This is the most-debated part of the rollout on X. OpenAI's product intent with this update is that most users shouldn't need to manually pick a model — Sol (Plus/Pro) or Luna (Free/Go) is meant to handle the Instant-vs-reasoning tradeoff automatically, with the new reasoning-effort slider (Plus/Pro) or "Think" button (Free/Go) as the remaining manual control for cases where a user wants more deliberate reasoning on demand.

The pushback: some users report the direct model picker is no longer in the main chat composer and now requires going into settings each time to select a specific legacy model — a friction increase for power users who relied on quickly switching models mid-conversation. That's a legitimate usability complaint, not a factual dispute about what shipped; if you're a heavy model-switcher, expect to spend more clicks finding the option than you did before August 6.

Does this apply to the API, Codex, or ChatGPT Work?

No — this update is scoped to the everyday ChatGPT chat experience only, at launch. It does not extend to:

• ChatGPT Work — OpenAI's enterprise/team surface, covered separately in explainx.ai's ChatGPT Work vs Codex guide
• The API — developers building on GPT-5.6 Sol, Terra, or Luna via API continue to reference the pricing and access explainx.ai covered in the July 30 price-cut breakdown
• Codex — no changes to Codex's model routing or usage accounting from this announcement

Some users on X have already asked OpenAI to extend unlimited Luna access to Codex as well — a reasonable ask given OpenAI's own Codex usage-limit changes throughout July — but nothing in the August 6 announcement commits to that. If it happens, it will be a separate announcement.

The reasoning-effort slider and the "Think" button

Two distinct controls ship with this update, split by plan tier:

The design logic is consistent with what OpenAI has said about GPT-5.6 since its June preview: durable capability tiers (Sol, Terra, Luna) that advance on their own cadence, with the user-facing complexity concentrated in how much reasoning a model applies rather than which model to pick. It's the same instinct behind Sol's max reasoning effort and Ultra mode at the API level, now surfaced as a simpler slider for chat.

What OpenAI's leadership said

Sam Altman (@sama):

"5.6 Sol much better in chat now and unlimited text chat for free users!"

Greg Brockman (@gdb), on the broader product philosophy behind the change:

"working hard to empower everyone with intelligence that works for them, in the form of great products which get simpler to use over time"

Brockman also called Luna "a quite capable model" — notable framing given Luna is the tier now carrying unlimited free-tier chat traffic rather than being positioned purely as a cost-optimized bulk tier, which is how it launched in June.

How this fits the rest of 2026's GPT-5.6 story

This is the fourth major GPT-5.6-related update explainx.ai has covered since the family's June 26 preview:

1. June 26 — GPT-5.6 Sol, Terra, Luna preview, limited partner access
2. July 9 — Public GA rollout in ChatGPT, Codex, and API
3. July 30 — Luna price cut 80%, Terra 20%, Sol Fast mode
4. August 6 — This update: Sol unifies Instant/reasoning for Plus/Pro; Luna goes unlimited for Free/Go

The pattern across all four: OpenAI keeps making the top of the funnel (Free, Go, Luna) meaningfully more capable and more accessible, while holding Sol's core pricing and access model steady for paying and developer tiers. Compare that trajectory against Claude Fable 5's own access and pricing timeline — both labs spent the back half of 2026 competing as much on who gets access to what, for how much as on raw benchmark scores. See explainx.ai's GPT-5.6 vs Claude Fable 5 comparison for the fuller tier-by-tier picture this update doesn't change.

Honest limitations

• The 68% figure is OpenAI's own benchmark, not third-party verified — treat it as directional, the same caveat that applies to every self-reported factuality or benchmark claim in 2026.
• "Unlimited" is scoped to text chat for Free/Go users — OpenAI's announcement doesn't cover voice, image generation, or other features.
• The model picker complaint is real — some users have lost the quick, in-composer way to switch models and now have to dig into settings, a genuine regression for power users even if it's an improvement for everyone else.
• This doesn't touch Codex, the API, or ChatGPT Work — don't assume enterprise or developer accounting changed based on this announcement alone.
• Rollout is gradual — "starting tomorrow" (August 7) for Free/Go unlimited chats means it may not be live on every account the instant this post publishes; check your own ChatGPT app if you don't see it yet.

What to check in your account

1. Plus/Pro users — look for the new reasoning-effort slider in the chat composer; if you don't see it yet, the rollout may still be reaching your account.
2. Free/Go users — check whether text chat shows a usage cap; if unlimited hasn't landed by August 8, it's likely still in the rollout wave.
3. Power users who switch models often — the direct picker has moved into settings; budget the extra click if you relied on fast switching.
4. Developers and enterprise teams — no action needed; this update doesn't change API, Codex, or ChatGPT Work behavior.
5. Anyone evaluating the factuality claim — test it against your own finance, medicine, or legal use case rather than taking the 68% figure as a guarantee for your workload.

Related on explainx.ai

• GPT-5.6 Sol, Terra, Luna — original June 2026 preview and July 9 GA
• GPT-5.6 vs Claude Fable 5 — complete tier comparison
• OpenAI cuts GPT-5.6 Luna 80%, Terra 20% (July 30 update)
• GPT-5.6 Sol usage limits — 18% increase, 5-hour return (July 2026)
• ChatGPT Work vs Codex — complete guide
• The true cost of AI subscriptions — ChatGPT, Claude, Gemini
• How to read AI benchmarks — a practical guide
• AI benchmarks — complete guide 2026
• GPT Live — ChatGPT voice, Go/Plus/Pro default (July 2026)
• Claude usage limits — 2026 timeline explained

Official sources: @OpenAI announcement · @sama on X · @gdb on X

Details reflect OpenAI's August 6, 2026 announcement (~5pm PT) and public statements from Sam Altman and Greg Brockman as of publication. The Free/Go unlimited-chat rollout was announced as starting August 7, 2026, and may reach accounts gradually — verify current availability inside your own ChatGPT app.