OpenAI Launches GPT-5.6 With Three-Tier Pricing as GPT-4.5 Era Ends

OpenAI surprised the AI community on June 26, 2026 by unveiling GPT-5.6 -- a three-model family named Sol, Terra, and Luna where each tier targets a different price-to-performance point. The same day, the company confirmed that GPT-4.5 would be retired from ChatGPT on June 27, closing the GPT-4 chapter for good. Together, the two moves signal that tiered pricing -- one base-generation of intelligence offered at multiple cost brackets calibrated to workload complexity -- is now a deliberate and durable strategy from OpenAI, not just a one-off experiment.

What GPT-5.6 Sol, Terra, and Luna Actually Cost

OpenAI has structured GPT-5.6 around three distinct price bands, according to VentureBeat's coverage of the June 26 launch:

• Sol -- $5 per million input tokens, $30 per million output tokens
• Terra -- $2.50 per million input tokens, $15 per million output tokens
• Luna -- $1 per million input tokens, $6 per million output tokens

The spread is striking: Sol costs five times more per output token than Luna. Terra is positioned as the middle ground -- reportedly matching GPT-5.5 performance at roughly half the cost, making it the clearest migration path for teams already running on GPT-5.5.

Actual API access remains limited at launch. OpenAI shared the models with the U.S. government before public rollout, and general availability across the API is expected "in the coming weeks." Around 20 organizations have access now.

What Each GPT-5.6 Tier Is Built For

Sol is the top-of-range option: designed for complex multi-step reasoning, extended coding sessions, security-sensitive workloads, and agentic pipelines where getting the answer right on the first pass matters more than cost per call. Think legal analysis, code generation for production systems, or agent loops that synthesize large document sets.

Terra is the workhorse. Competitive benchmark performance at mid-range pricing makes it the rational default for most enterprise use cases -- especially for teams whose workflows already run reliably on GPT-5.5 and want to cut spend without changing behavior.

Luna is OpenAI's direct entry into the sub-$2 input tier now occupied by models like Gemini 3.5 Flash ($1.50/M input). High-volume classification, summarization, simple Q&A -- tasks where quality is good enough and you are calling the model millions of times per day. Luna's pricing is the lowest OpenAI has publicly offered for a GPT-5-generation model.

GPT-4.5 Retires Today, Closing the GPT-4 Chapter

June 27, 2026 is GPT-4.5's final day in ChatGPT. OpenAI confirmed the removal follows a 30-day sunset period; the stated rationale is that GPT-5 and GPT-5.3-Codex have fully superseded GPT-4.5 on every major benchmark. API developers are unaffected -- the GPT-4.5 API was already discontinued in July 2025, more than a year earlier.

The retirement is symbolic as much as operational. GPT-4 launched in March 2023 at $30 per million input tokens. Three years later, OpenAI is launching the lowest-tier GPT-5.6 model at $1 per million input tokens -- a 30x input price reduction at the entry level across one generational arc. That compression is real, and it has largely been passed to developers rather than absorbed as margin.

Why Three Tiers From One Provider Changes the Math

A single-model lineup is simple: you use the model or you do not. Three tiers from the same provider -- at 5x output price differences -- means the correct choice shifts query by query. A knowledge assistant that calls Sol for every message is paying five times the cost it would for routine questions that Luna handles equally well.

This fragmentation already existed across providers (Claude vs. Gemini vs. GPT), but GPT-5.6 brings it inside a single vendor family. The practical consequence is that manual tier selection becomes increasingly unworkable at scale. Automated routing -- matching each query to the cheapest model capable of answering it -- becomes more valuable as provider menus grow wider. ByteChat is one tool built on this premise: bring your own keys across providers, route based on capability, and pay the lab directly with zero markup.

Frequently Asked Questions

What is GPT-5.6 and when can I access it?

GPT-5.6 is OpenAI's newest model family, announced June 26, 2026, with three tiers -- Sol, Terra, and Luna -- targeting complex, balanced, and high-volume workloads respectively. General API availability is expected within weeks; as of launch, access is limited to roughly 20 partner organizations that OpenAI briefed alongside the U.S. government.

How does GPT-5.6 pricing compare to GPT-5.5?

GPT-5.6 Terra ($2.50/$15 per million tokens) is approximately half the cost of GPT-5.5 ($5/$30 per million tokens) while reportedly matching its performance. GPT-5.6 Luna ($1/$6) is the cheapest GPT-5-generation model OpenAI has publicly offered.

What should I use now that GPT-4.5 is retired from ChatGPT?

For most ChatGPT users, the default model (currently GPT-5.5) is the practical successor. API developers were already migrated away from GPT-4.5 in mid-2025; those users will find GPT-5.6 Terra the closest equivalent on performance-per-dollar once general access opens.

The GPT-4 chapter is officially closed -- the question now is which tier of its successor your workload actually needs.