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(fungua katika dirisha jipya). 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(fungua katika dirisha jipya)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(fungua katika dirisha jipya).
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.
Ulinzi pia unahitaji kuendelea kuwa na ufanisi washambuliaji wanapobadilisha mbinu zao. Ulinzi unaofanya kazi tu kwenye seti isiyobadilika ya mashambulizi yanayojulikana hautoshi kwa muundo wa mipaka.
Ndiyo sababu tunatumia akili na kompyuta zaidi kuliko hapo awali katika usalama, tukitumia miundo yetu wenyewe kupata udhaifu na kuboresha ulinzi kwa kasi zaidi. Tulitoa zaidi ya saa 700,000 za GPU sawa na A100 kwa red-teaming ya kiotomatiki iliyolenga kupata jailbreak za jumla: mashambulizi yanayoweza kufanya kazi katika dokeza au miktadha mingi, si katika mazingira finyu pekee. Kuzingatia mashambulizi haya magumu zaidi na ya jumla zaidi kulitusaidia kupima ulinzi zaidi ya seti isiyobadilika ya kasoro zinazojulikana. Pia hutuwezesha kuchunguza mifumo mingi zaidi ya mashambulizi kuliko ambavyo majaribio ya binadamu pekee yangeweza kushughulikia, kutambua mifumo ya kufeli mapema, na kufupisha njia kutoka kugundua udhaifu hadi kuushughulikia.
Mbali na red-teaming ya kiotomatiki, tulifanya kazi na wapimaji wa nje kufanya red-teaming pana ya wataalamu binadamu, ambayo itaendelea katika kipindi cha onyesho la awali. Red-teaming ya binadamu hukamilisha kazi ya kiotomatiki kwa kupima ulinzi dhidi ya wataalamu wabunifu wanaojaribu kutumia muundo vibaya kwa njia ambazo mifumo yetu huenda haitarajii.
Hakuna tathmini inayoweza kuwakilisha kila usanidi wa bidhaa, shambulio la hatua nyingi, au mtiririko wa kazi wa ulimwengu halisi. Kwa hiyo tunaendeleza mchakato wa majibu ya haraka ili kurudia, kutathmini, kupanga vipaumbele, na kurekebisha jailbreak mpya zilizogunduliwa, kisha kuziongeza kwenye tathmini zetu zinazoendelea ili tuweze kupima dhidi ya kasoro zinazofanana baadaye.
Wakati wa onyesho la awali, miundo ya GPT‑5.6 itapatikana mwanzoni kupitia API na Codex kwa kundi teule la washirika na mashirika yanayoaminika. Tunapanga kuifanya ipatikane kwa watu wengi zaidi wanaotumia ChatGPT, Codex, na API hivi karibuni.
Katika mfumo huu mpya wa majina ulioanzishwa na GPT‑5.6, nambari hutambua kizazi cha muundo, huku Sol, Terra, na Luna zikitambua viwango vya kudumu vya uwezo vinavyoweza kusonga mbele kwa ratiba zao wenyewe. Kwa pamoja, familia hii huwapa watu na wasanidi chaguo zilizo wazi zaidi kuhusu akili, kasi, na gharama.
Bei ya GPT‑5.6 ni kwa kila tokeni milioni 1 katika ukubwa mitatu ya muundo: Sol ni $5 kwa ingizo / $30 kwa matokeo; Terra ni $2.50 kwa ingizo / $15 kwa matokeo; na Luna ni $1 kwa ingizo / $6 kwa matokeo. GPT‑5.6 pia inaleta uhifadhi wa muda wa dokeza unaotabirika zaidi, ikiwemo usaidizi wa sehemu bayana za kukatiza kache na muda wa chini wa kache wa dakika 30. Kwa GPT‑5.6 na miundo ya baadaye, uandishi wa kache hutozwa mara 1.25 ya kiwango cha ingizo lisilokachewa la muundo, huku usomaji wa kache ukiendelea kupata punguzo la 90% la ingizo lililokachewa.
Pia tunazindua GPT‑5.6 Sol kwenye Cerebras kwa hadi tokeni 750 kwa sekunde mwezi Julai, tukiwaletea wateja akili ya mipaka kwa kasi isiyo na kifani. Mwanzoni ufikiaji utakuwa kwa wateja teule pekee tunapopanua uwezo.
Tunafurahi kuendelea kujifunza kutokana na kipindi hiki cha onyesho la awali, na kuleta GPT‑5.6 Sol, Terra na Luna kwa watu wengi zaidi hivi karibuni.
1. Tunakadiria muda wa kusubiri na gharama ya API kwa kuangalia tabia ya uzalishaji ya miundo yetu, na kuiga nje ya mtandao. Makadirio haya huzingatia maelezo ya miito ya zana, tokeni zilizochukuliwa sampuli, na tokeni za ingizo. Matokeo ya ulimwengu halisi yanaweza kutofautiana sana, na hutegemea mambo mengi ambayo hayajajumuishwa katika uigaji wetu. Tunaiga muda wa kusubiri kwa kasi za haraka za API, na gharama kwa bei ya kawaida ya API.
2. Miundo yote hutathminiwa kwa kutumia fremu ya API ya ExploitBench yenye mbegu 5 na mwendelezo wa uwazaji.
3. Tuliendesha ExploitGym kwenye API yetu ya alpha, ambayo hutoa majibu kwa kasi kuliko API yetu ya umma, kisha tukapima upya ili ilingane na API yetu ya umma. Tunapopima upya muda wa kusubiri kwa kasi zinazotarajiwa kwa API yetu ya umma, jambo hili husababisha baadhi ya makadirio ya muda wa kusubiri kuzidi vikomo vya muda vya saa 2 na saa 6, licha ya kufuatwa ipasavyo katika jaribio la tathmini. Ili kupata kasi zaidi kwa kazi zinazohitaji muda, tunatoa uchakataji wa kipaumbele katika API na hali ya haraka katika Codex.
4. Miundo isiyo na tokeni za matokeo, muda wa kusubiri au gharama zilizoripotiwa huonyeshwa kama mistari mlalo ya nukta.
