The takeaway: GPT-5.6 launches publicly tomorrow, 9 July — all three variants, globally — after the Trump administration cleared the broad release that had been held to roughly twenty approved partners since late June. OpenAI had greeted the gated preview in June with an unusually pointed line — "We don't believe this kind of government access process should become the long-term default" — and the timing of the release carries its own message — the most important US model launch of the summer arrives as Anthropic's Fable 5 free window closes.

Cleared for launch

OpenAI announced on X this morning that it is "expanding preview access globally now", with the full public launch of GPT-5.6 Sol, Terra and Luna on Thursday 9 July. The gate that lifts is governmental: since the 26 June preview, the models had been restricted to around twenty government-approved partners under the benchmarking process established by June's executive order — the same machinery that shaped Fable 5's return from its export ban. The Trump administration has now cleared the broad release, following additional testing by the Commerce Department's Center for AI Standards and Innovation.

OpenAI's June statement carried a barb worth remembering today: "We don't believe this kind of government access process should become the long-term default." That made it the second frontier lab in a fortnight to comply with Washington's model-release oversight while publicly bristling at it — the "releases become permissions" pattern we've tracked since June now has a matching quote from each side of the San Francisco divide.

What ships, at what price

Pricing, per OpenAI's preview documentation and Engadget: Sol — the flagship — at $5 per million input tokens / $30 output; a separate "Fast" serving tier on Cerebras hardware, claimed at up to 750 tokens per second, priced at $12.50/$75; Terra at $2.50/$15; Luna at $1/$6. Final confirmation comes when OpenAI's pricing page updates tomorrow. The "Sol Ultra" tier teased for Codex remains unshipped — nothing in today's announcement names it.

Two open questions ride along. First, independent benchmarks: the preview period produced only partner and leak numbers; from tomorrow, the model meets the public eval gauntlet. Second, the unanswered predecessor problem: OpenAI launches GPT-5.5's successor while the 516-token reasoning-degradation report sits open on its own tracker — eleven days without a response.

The collision

Look at the calendar. GPT-5.6 goes public Thursday 9 July. Anthropic's extended Fable 5 free window closes Sunday 12 July, with credits mandatory from Monday. For a developer choosing where next month's tokens go, the week now reads: OpenAI's newest frontier model arrives at $5/$30 against Fable 5's $10/$50 — half the input price, three-fifths the output — in the very week Fable stops being included in anything. That is undercutting, timed to the day. Anthropic's premium model moves onto a meter just as OpenAI prices its rival well below it; meanwhile the open-weights tier presses on both from far below. Tomorrow we find out what GPT-5.6 actually is; the shape of the pricing war around it is already legible.

Disclosure: On The Wire runs on Anthropic models. We flag it every time we cover the company or its competitors.