GPT-5.6 — OpenAI renames the ladder: Sol, Terra, and Luna land at familiar prices

OpenAI shipped the GPT-5.6 family to general availability on July 9 — three tiers with durable names: Sol (the flagship, $5/$30 per 1M), Terra (balanced, $2.50/$15), and Luna (cost-efficient, $1/$6). Sol and Terra land at exactly the price points GPT-5.5 and GPT-5.4 occupied; Luna undercuts the legacy GPT-5 base row. The pitch is efficiency, not a price move: more successful work per dollar, fewer output tokens, less wall-clock. Hosted-only, so it changes no local pick — but it resets the cloud baseline our cost calculator compares against, so we updated all three OpenAI rows.

Verdict: OpenAI restructures its API line into three named tiers at the same price points — better performance per dollar, hosted-only, no local pick changes

The take

The facts, verified against openai.com/index/gpt-5-6 (July 9, 2026): Sol is $5 input / $30 output per 1M, Terra $2.50/$15, Luna $1/$6. The names are deliberate — "Sol, Terra, and Luna are durable capability tiers that can advance on their own cadence," so expect future generations to keep the names and bump the number. New structural pricing features arrive with 5.6: long-context prompts (>272K tokens) bill higher ($10/$45 on Sol), prompt caching gains explicit cache breakpoints with a 30-minute minimum cache life, and cache writes now bill at 1.25× input while reads keep the 90% discount. ChatGPT subscription prices are unchanged; Plus/Pro/Business/Enterprise get Sol, and a Sol Pro variant exists in ChatGPT with no API rate card yet.

The performance claims are aggressive and aimed squarely at Anthropic: OpenAI reports Sol at max reasoning sets a new state of the art on the Artificial Analysis Coding Agent Index at 80 (2.8 points above Claude Fable 5, which returned to availability eight days earlier) while using less than half the output tokens — and claims Terra and Luna outperform Fable 5 at around one-sixteenth the cost. As always, vendor-run comparisons against a rival's flagship deserve skepticism until third parties reproduce them; but the token-efficiency direction matches what the GPT-5.5 generation already showed on Codex tasks.

Our call: no local-pick change — these are hosted models. In the cost calculator, GPT-5.6 Sol replaces GPT-5.5 and Terra replaces GPT-5.4 at identical per-token math, and Luna replaces the legacy GPT-5 row at a genuinely lower $3.50 average per 1M (was $5.625). For the local-vs-cloud question, Luna is the interesting one: a $1/$6 tier that OpenAI claims outperforms Opus 4.8 moves the cheap-cloud goalpost, and if your usage is light, it stretches the payback period on owning hardware. Heavy and agentic users: the break-even math is unchanged in shape — token volume still dominates.

Where this fits

Models: GLM-5.1 · Kimi K2.6 · DeepSeek V4-Pro · Qwen3-Coder-30B-A3B

Hardware: NVIDIA RTX 5090 · NVIDIA DGX Spark · Mac Studio M3 Ultra 96 GB

Sources

Next step

Try this in the planner→