Today OpenAI is previewing GPT-5.6 Sol, its newest flagship model for developers and enterprises, alongside Terra and Luna.

Sol is built for frontier reasoning and long-horizon agentic work; Terra is a balanced everyday model with GPT-5.5-competitive performance at 2x lower cost; and Luna is the fastest, most affordable member of the family.

Breaking new ground

GPT-5.6 Sol advances coding, scientific reasoning, long-horizon planning, and agentic workflows, while improving reliability and efficiency across demanding real-world tasks. Sol establishes new high-water marks across some of OpenAI’s most challenging evaluations:

• Coding: GPT-5.6 Sol sets a new SOTA on Terminal-Bench 2.1, which tests command-line workflows requiring planning, iteration, and tool coordination.

• Cybersecurity: GPT-5.6 Sol shifts the efficiency frontier for vulnerability research and controlled exploitation tasks, achieving competitive results on ExploitBench while using roughly one-third of the output tokens compared with another leading frontier system.

• Biology: On SecureBio evaluations, GPT-5.6 reached top reported scores including 53.5% on the Virology Capabilities Test, 60.0% on Molecular Biology, 68.4% on Human Pathogen Capabilities, and 68.3% on World-Class Bio, about 9 percentage points above GPT-5.5.

These gains are especially important for developers building agents that need to reason over large codebases, operate tools, debug multi-step failures, conduct research, or assist with defensive security work.

What’s new

GPT-5.6 introduces a new max reasoning effort, giving Sol more time to reason deeply on difficult tasks. It also introduces ultra mode, which goes beyond a single-agent setup by using subagents to accelerate complex work.

During the preview, GPT‑5.6 models will initially be available through the API and Codex to a select group of trusted partners and organizations. OpenAI plans to make them more broadly available to people using ChatGPT, Codex, and the API soon.

Learn more

I just read that the US gov wants to bless 5.6 before release. Is this a tamed down version?

Edit: Nevermind - just got the answer

I loved the new naming pattern, gives them a little bit more of personality.

So for Pro models would be Nova or similar? Just wondering

TL;DR: OpenAI is previewing GPT-5.6 in three variants (Sol, Terra, and Luna), with Sol delivering state-of-the-art reasoning, coding, science, and cybersecurity performance, slightly outperforming Claude Mythos 5 while using roughly 80% fewer output tokens on ExploitBench. OpenAI is initially rolling it out to a limited group while completing a temporary U.S. government AI safety review before broader availability.

I´m so curious about these variants of GPT-5.6 and I can´t wait to work with it.

I’m curious too, also how the newer models will integrate in production environments. Hope that gets super fast!

Great news, looking forward to it.

I think one can infer that, by pricing 2x between gpt-5.4 and gpt-5.5, the delivery of gains within a month was likely by a 2x test-time compute, technology like “pro” does with 12x the billing - and where the pro per-token bill was not increased between these version steps.

So I would guess we have similar (gpt-5.4->gpt-5.5) === (gpt-5.6 terra->gpt-5.6 sol) - at least as your pricing guidance.

Notable is API pro has no context window cache discount, so that can make costs of tool iterating extreme.

how do i get enrolled in early access without being a f500 or famous person? the way i use these models for work and open source prototypes is sufficiently out of distribution in terms of software design and style that i have to basically red team the ethically and moral reasoning of an llm to evaluate how logically coherent they are for ood work

cartazio:

enrolled in early access

Enjoy the only document needed - that partners invited to preview are on the level of having their own account rep to help them through required documents and government reporting:

https://help.openai.com/en/articles/20001325-a-preview-of-gpt-56-sol-terra-and-luna

You can also enjoy this dark theme background color rotation while you wait.

TOKEN_HIGHLIGHT_COLORS: tuple[str, str, str] = (
    "#55291d",  # rust-orange
    "#3d2366",  # violet-purple
    "#0c433d",  # green-cyan
)

The wording “more broadly available” does not sound promising.

It will be interesting to see how this rollout plays out, especially for non-US residents.

The timing on this couldn’t be better.

I run agentic systems daily - OpenClaw, Hermes, Claude Code orchestrating multiple AI workers.

The bottleneck has always been cost at scale.

Anthropic’s API pricing makes it brutal to run agents 24/7. You’re watching credits evaporate in real time.

The fact that OpenAI allows third-party harnesses to tap into these models through an existing subscription changes the math completely.

Looking forward to Sol Ultra powering my agents without per-token anxiety.

And “Ultra” mode with subagents working together - that’s exactly where agentic AI needs to go.

Thank you for making this accessible to builders, not just enterprises with infinite API budgets.

Time to put these through their paces.

I’ve got 6 DGX Sparks running great local model like Gemma4 and these 5.6 models are going to run it all.

Welcome to the community, @JeffJHunter … hope you stick around. We’ve got a wealth of information and a great group of people here!

The industry keeps acting like the model is the product. Cute.

A model is compute. The actual system is external state, memory, governance, and persistence. If all of that lives inside the model, you are not designing architecture. You are just overloading the poor thing with bells and whistles while hoping for something magical too happen.

Pimpcat:

Cute

Fortunately we can follow your strange complaint’s logic:

If (all of that lives inside the model)

as a conditional statement evaluates to falsey. Thus, we’ve short-circuited the rest of your post to a no-op.

For Codex users the model pricing for GPT 5.5 was listed as Credits

Based on 5.6 release notes I see there is only the info about pricing in $ . do we know whether there is a shift? or later on the Credit mapping for Codex will be posted later on?