We're beginning a limited preview of the GPT‑5.6 series: Sol, our flagship model; Terra, a balanced model for everyday work; and Luna, a fast and affordable model. Terra has competitive performance to GPT‑5.5 while being 2x cheaper and Luna brings strong capability at our lowest cost.
GPT‑5.6 Sol launches with our most robust safety stack to date. We strengthened protections for higher-risk activity, sensitive cyber requests, and repeated misuse, and spent multiple weeks finding weaknesses, pressure-testing our system, and hardening it against real-world attacks.
We believe in broad access, and we plan to make GPT‑5.6 Sol, Terra, and Luna generally available in the coming weeks. As part of our ongoing engagement with the U.S. government, we previewed our plans and the models’ capabilities ahead of today’s launch. At their request, we are starting with a limited preview for a small group of trusted partners whose participation has been shared with the government, before releasing more broadly. During this preview, we will continue testing and coordinating closely with partners as we work toward broader availability. We don’t believe this kind of government access process should become the long-term default. It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them. We are taking this short-term step because we believe it is the strongest path to broader availability in the coming weeks, while we work with the Administration to develop the cyber Executive Order framework and a repeatable process for future model releases.
GPT‑5.6 Sol is our strongest model yet. To give a preview of model performance, we share a set of evaluations highlighting improved agentic capabilities in coding, biology, and cybersecurity, with additional safety and preparedness evaluations available in our system card(osclaíonn i bhfuinneog nua). We will share an expanded suite of evaluation results when we make the model broadly available.
With GPT‑5.6, we’re introducing a new max reasoning effort to give Sol the most time to reason deeply. Additionally, we’re introducing a new ultra mode that goes beyond the capabilities of a single agent by leveraging subagents to accelerate complex work.
For coding workflows, GPT‑5.6 Sol sets a new state of the art on Terminal‑Bench 2.1, which tests command-line workflows requiring planning, iteration, and tool coordination.
GPT‑5.6 Sol also shows broad improvements in biology workflows. On GeneBench v1, which evaluates long-horizon genomics and quantitative-biology analyses, it achieves stronger results than GPT‑5.5 while using fewer tokens.
GPT‑5.6 Sol is our most capable model yet for cybersecurity. It shifts the performance-efficiency frontier for long-horizon security tasks including vulnerability research and exploitation. On ExploitBench², GPT‑5.6 Sol is competitive with Mythos Preview using only ~1/3 of the output tokens. On ExploitGym(osclaíonn i bhfuinneog nua)3, a benchmark created by UC Berkeley researchers in collaboration with OpenAI and other frontier labs, GPT‑5.6 Sol, Terra, and Luna models all demonstrate strong improvements in cyber capabilities as we increase reasoning.
We developed GPT‑5.6 Sol, Terra and Luna with our most robust safeguards to date, with configurations matched to each model’s capabilities. As the model becomes more capable, we design safeguards to increasingly hold up to real-world adversarial pressure while preserving access to legitimate work such as code review, vulnerability research, patch development, debugging, security education, and defensive testing. Our goal is to make prohibited offensive activity more difficult, uncertain, and detectable without unnecessarily limiting those beneficial uses. Based on our assessment of the model and safeguards, we expect substantial benefit for legitimate defensive work, while meaningfully constraining prohibited offensive use.
GPT‑5.6 Sol is better at helping people find and fix vulnerabilities than reliably carrying out end-to-end attacks. As these capabilities continue to advance, our priority is to make sure they reach and benefit defenders, who can use these tools to find weaknesses, develop patches, and strengthen systems more broadly.
GPT‑5.6 Sol does not cross the Cyber Critical threshold under our Preparedness Framework. In evaluations involving Chromium and Firefox, it identified bugs and exploitation primitives—the building blocks of an exploit—but did not autonomously produce a functional full-chain exploit under the conditions tested. Still, benchmark thresholds cannot capture every way a model may be used or combined with other tools. That uncertainty, along with the model’s broader step change in capabilities, is why we are pairing the model’s increased capabilities with stronger safeguards and a phased release. We share more details about our safeguards in the GPT‑5.6 Preview system card(osclaíonn i bhfuinneog nua).
No single safeguard is sufficient against determined or adaptive misuse. Across the GPT‑5.6 preview, we use layered safeguards, with exact configurations varying across models, and pressure-test them for real-world attacks. These include protections trained into the model, real-time checks during generation, account-level signals, differentiated access, monitoring, enforcement, and continued testing.
GPT‑5.6 is trained to refuse prohibited cyber assistance, including when users attempt to disguise their intent or jailbreak the model. These model-level safeguards establish the first boundary around what the model should and should not help with.
Real-time cyber and biology misuse classifiers provide another layer by evaluating output as it is generated. For higher risk cases, if they detect a potential violation, the generation may be paused while a larger reasoning model reviews the conversation and its context. If the output is assessed as disallowed, it is withheld before it reaches the user.
Flagged activity can also trigger account-level review across relevant conversations and risk signals, consistent with our terms and policies around content retention and review. Looking beyond a single conversation helps our systems distinguish persistent malicious behavior from legitimate dual-use security work, where similar technical concepts may appear in very different contexts.
Together, these layers make the overall approach more robust than any one safeguard on its own. Model behavior reduces the likelihood of harmful responses, real-time systems can intervene during generation, account-level review can identify broader patterns, and differentiated access preserves important defensive work without making the most sensitive capabilities broadly available by default.
Especially during the preview, users may encounter safeguards that block or refuse some requests. Other requests may take longer because generation is paused for additional review. Safeguards may occasionally intervene on legitimate work, particularly in dual-use areas where defensive and offensive activity can initially look similar.
That is part of what the preview is designed to test. We want to understand not only whether the safeguards constrain misuse, but whether legitimate users can still complete normal work reliably and efficiently. Feedback during the preview will help us reduce unnecessary blocks and delays, improve how the safeguards interpret context, and create a smoother experience before wider release.
We are also working with enterprise customers on longer-term approaches—including privacy-preserving detection, customer-operated safety controls, and access calibrated to the risk of a customer, user, or workload—to advance safety while supporting enterprise privacy requirements.
Ní mór do chosaintí fanacht éifeachtach freisin nuair a athraíonn ionsaitheoirí a gcuid beart. Ní leor do shamhail ar thús cadhnaíochta cosaint nach n-oibríonn ach ar shraith sheasta ionsaithe aitheanta.
Sin é an fáth a bhfuil níos mó intleachta agus ríomhaireachta ná riamh á gcur i bhfeidhm againn ar shábháilteacht, ag úsáid ár samhlacha féin chun laigí a aimsiú agus cosaintí a fheabhsú níos tapa. Chaith muid breis agus 700,000 uair GPU de choibhéis A100 ar mhodh na foirne deirge uathoibrithe chun sáruithe srianta uilíocha a aimsiú: ionsaithe a d’fhéadfadh oibriú thar go leor leid nó comhthéacsanna, ní hamháin i suíomh cúng amháin. Trí dhíriú ar na hionsaithe níos deacra agus níos ginearálta seo, bhíomar in ann na cosaintí a thástáil thar shraith sheasta teipeanna aitheanta. Ligeann sé dúinn i bhfad níos mó patrún ionsaithe a fhiosrú ná mar a chlúdódh tástáil dhaonna amháin, patrúin teipe a aithint níos luaithe, agus an bealach ó laige a aimsiú go dtí í a réiteach a ghiorrú.
Chomh maith le modh na foirne deirge uathoibrithe, d’oibríomar le tástálaithe tríú páirtí chun modh na foirne deirge daonna ó shaineolaithe a dhéanamh, rud a leanfaidh ar aghaidh le linn na tréimhse réamhamhairc. Comhlánaíonn tástáil dhaonna modh na foirne deirge an obair uathoibrithe trí chosaintí a thástáil i gcoinne saineolaithe cruthaitheacha atá ag iarraidh an tsamhail a mhí-úsáid ar bhealaí nach mbeadh ár gcórais ag súil leo.
Ní féidir le haon mheastóireacht gach cumraíocht táirge, ionsaí ilchéime ná sreabhadh oibre sa saol fíor a léiriú. Dá bhrí sin, coinnímid próiseas mearfhreagartha chun sáruithe srianta nua-aimsithe a atáirgeadh, a mheas, a chur in ord tosaíochta agus a leigheas, agus ansin iad a chur lenár measúnuithe leanúnacha ionas gur féidir linn tástáil a dhéanamh ar theipeanna cosúla amach anseo.
Le linn an réamhamhairc, beidh samhlacha GPT‑5.6 ar fáil ar dtús tríd an API agus Codex do ghrúpa roghnaithe comhpháirtithe agus eagraíochtaí iontaofa. Tá sé beartaithe againn iad a chur ar fáil níos forleithne go luath do dhaoine a úsáideann ChatGPT, Codex agus an API.
Sa chóras ainmniúcháin nua seo a tugadh isteach le GPT‑5.6, léiríonn an uimhir ghlúin na samhla, agus léiríonn Sol, Terra agus Luna sraitheanna buana cumais ar féidir leo dul chun cinn ar a luas féin. Le chéile, tugann an teaghlach roghanna níos soiléire do dhaoine agus d’fhorbróirí ó thaobh intleachta, luais agus costais de.
Tá GPT‑5.6 praghsáilte in aghaidh 1M téacschomharthaí thar thrí mhéid samhla: Sol ag $5 ionchur / $30 aschur; Terra ag $2.50 ionchur / $15 aschur; agus Luna ag $1 ionchur / $6 aschur. Tugann GPT‑5.6 taisceadh leid níos intuartha isteach freisin, lena n-áirítear tacaíocht do bhrisphointí taisce follasacha agus saol íosta taisce 30 nóiméad. Do GPT‑5.6 agus samhlacha níos déanaí, gearrtar táillí ar scríbhinní taisce ag 1.25x ráta ionchuir neamhthaiscthe na samhla, agus leanann léamha taisce den lascaine 90% ar ionchur taiscthe a fháil.
Táimid ag seoladh GPT‑5.6 Sol ar Cerebras freisin ag suas le 750 comhartha sa soicind i mí Iúil, ag tabhairt intleacht ar thús cadhnaíochta do chustaiméirí ar luas nach raibh a leithéid ann riamh cheana. Beidh rochtain teoranta ar dtús do chustaiméirí roghnaithe agus muid ag leathnú acmhainne.
Táimid ar bís leanúint ag foghlaim ón tréimhse réamhamhairc seo, agus GPT‑5.6 Sol, Terra agus Luna a thabhairt chuig níos mó daoine go luath.
1. Measaimid moill agus costas API trí bhreathnú ar iompar táirgthe ár samhlacha agus trí ionsamhlú as líne. Cuireann na meastacháin seo sonraí glaonna uirlise, téacscomharthaí sampláilte agus téacscomharthaí ionchuir san áireamh. D’fhéadfadh torthaí sa saol fíor athrú go mór agus braitheann siad ar go leor tosca nach bhfuil gafa inár n-ionsamhlú. Ionsamhlaímid moill ag luasanna tapa API, agus costas ag gnáthphraghsáil API.
2. Déantar gach samhail a mheas trí leas a bhaint as creat API ExploitBench le 5 shíol agus leanúnachas réasúnaíochta.
3. Rith muid ExploitGym ar ár API alfa, a aschuireann freagraí níos tapúla ná ár API poiblí, agus ansin rinneamar athscálú chun ár API poiblí a mheaitseáil. Nuair a athscálaítear moilleanna chuig na luasanna a bhfuiltear ag súil leo dár API poiblí, cuireann sé seo ar roinnt moilleanna measta dul thar na teorainneacha ama 2u agus 6u, cé gur cloíodh leo i gceart sa rith mheastóireachta. Chun luasanna níos tapúla a fháil d’obair atá íogair ó thaobh ama de, cuirimid próiseáil tosaíochta ar fáil san API agus mód tapa in Codex.
4. Léirítear samhlacha nach bhfuil téacscomharthaí aschuir, moill ná costas tuairiscithe acu mar línte cothrománacha poncaithe.
