GPT 5.6 release

GPT-5.6 is slated for general (and international) release on July 9. The model family comprises of three sizes: Sol, Terra and Luna. The model card is here. A “select group of trusted partners and organizations” had preview access, including Ethan Mollick, Derya Unutmaz and Peter Gostev.

Ethan Mollick@emollick

I was an early tester of GPT-5.6 Sol. I was asked to not share demos until after launch but it is a very good model. It is of similar ability, but quite different feel, than Fable. Fable wants to go off and do work on its own pace, Sol is faster but works with you in steps more.

4:30 AM · Jul 8, 2026 · 283K Views

---

105 Replies · 135 Reposts · 3.28K Likes

GPT-5.6 is said to “feel” part of the GPT-5 family while targeted against Anthropic’s Fable 5 as an intermediate response before GPT-6 (pre-)release.

Ethan Mollick@emollick

Fable feels very different than Opus. GPT-5.6 feels like a part of the GPT-5 family. I developed a very complex set of heuristics about when to use which. Fable was often “smarter” but was also too self-directed for some work, while that characteristic was perfect for others.

6:06 AM · Jul 8, 2026 · 26.3K Views

---

7 Replies · 10 Reposts · 348 Likes

Mitchell Hashimoto@mitchellh

I had early access to 5.6/Sol for ~month. Sol is my default. It is faster, plans/judges just as good as Fable, and I think produces better overall work. I’ll reach for Fable still for highly targeted debug or performance work with clear reward functions. A cheeky way I describe

2:27 PM · Jul 8, 2026 · 172K Views

---

134 Replies · 146 Reposts · 3.59K Likes

GPT-5.6 Sol improves on hallucinations, where as Terra and Luna fare worse than GPT-5.5.

GPT 5.6 Sol token prices remain the same as GPT-5.5, Luna remains more expensive than GPT-5.4 Mini:

GPT‑5.6 is priced per 1M tokens across three model sizes: Sol is $5 input / $30 output; Terra is $2.50 input / $15 output; and Luna is $1 input / $6 output. GPT‑5.6 also introduces more predictable prompt caching, including support for explicit cache breakpoints and a 30-minute minimum cache life. For GPT‑5.6 and later models, cache writes are billed at 1.25x the model’s uncached input rate, while cache reads continue to receive the 90% cached-input discount.

Prices for actual tasks could still vary, Artificial Analysis don’t have a report on that yet. A post by Peter Gostev suggests that GPT-5.6 might be cheaper:

Peter Gostev@petergostev

You'll get used to it quickly, but GPT-5.6 is sooo much more token efficient than 5.5. Going back to 5.5 was a massive pain and felt like going back a model generation.

3:18 PM · Jul 8, 2026 · 8.58K Views

---

13 Replies · 4 Reposts · 181 Likes

GPT-5.6 Sol will be available in Codex, but he advises caution on the reasoning levels and the fast option though:

Peter Gostev@petergostev

When you get access to GPT-5.6-Sol in Codex, be careful with how you are using your tokens. It is trivial to blow though your Pro, if you do everything & /fast via Max/Ultra with 10x sub-agents . Just be patient, do it when necessary, otherwise xHigh is a good default.

12:44 PM · Jul 8, 2026 · 34.4K Views

---

34 Replies · 11 Reposts · 494 Likes

---

Update 2026-07-09

• Availability

  • Sol available over the next 24 hours on paid plans

    • “the frontier model in the GPT-5.6 family. It roughly corresponds to the unsuffixed model tier used in earlier GPT-5 families”

  • Terra available on free plans

    • “designed for workloads that balance intelligence and cost. It roughly corresponds to the mini model tier used in earlier GPT-5 families.”

  • Luna also on Free

    • “Luna is designed for cost-sensitive, high-volume workloads. It roughly corresponds to the nano model tier used in earlier GPT-5 families.”

  • Also on the API: Model overview page and OpenRouter

    • not available as of yet on my (EU) account via OpenAI; only accessible via OpenRouter

  • Pro is the same model as Sol, with - in the API - reasoning.mode set to pro

    • there is a dedicated model entry on OpenRouter

• Performance

  • Saturates Prinzbench in ChatGPT, a privately withheld niche common law set of hard problems where GPT-5.5 Pro used to excel, Anthropic Fable 5 was behind and Sonnet 4.6 was last.

    

    prinz@deredleritt3r

    Some random cool things that GPT-5.6 Sol was able to do for me over the past few weeks: 1. Legal Research: prinzbench scores: GPT-5.4 Pro (extended): 79/99 GPT-5.5 Pro (extended: 82/99 GPT-5.6 Sol Pro: 91/99 As I wrote yesterday (and as you can see above), the Pro version of

    6:40 PM · Jul 9, 2026 · 5.35K Views

    ---

    14 Replies · 7 Reposts · 148 Likes

    Chart without GPT-5.6 Sol:

    

  • Sol is SoTA (and ahead of Fable) on TerminalBench 2.1, BrowseComp and Agents’ Last Exam

  • ArtificialAnalysis (blog post) places GPT-5.6 Sol second to Fable 5 (60 vs 59), but notes that it costs only a third (or half, depending whether you look at the total cost to run the benchmark or the “cost per task”)

Token Efficiency: the term used by Peter Gostev may mean that Sol is simply cheaper than its competitors:

Source: Launch Livesteam

• Model guidance document is here