OpenAI launches GPT-5.6 model family with Sol flagship

Investing.com -- OpenAI released its GPT-5.6 family of models for general availability on Thursday, following a limited preview period. The launch includes Sol, the company's new flagship model, alongside Terra, described as a balanced model for everyday work, and Luna, positioned as the most cost-efficient option.

GPT-5.6 Sol achieved a score of 53.6 on Agents' Last Exam, an evaluation of long-running professional workflows across 55 fields. This result surpassed Claude Fable 5 with adaptive reasoning by 13.1 points. At medium reasoning, Sol exceeded Fable 5 by 11.4 points at roughly one-quarter the estimated cost, according to OpenAI.

On the Artificial Analysis Intelligence Index, which measures intelligence spanning agentic work, coding, scientific reasoning, and general capabilities, GPT-5.6 Sol with max reasoning came within one point of Fable 5. The model completed tasks in 61% less time at roughly half the estimated cost.

For coding tasks, GPT-5.6 Sol with max reasoning scored 80 on the Artificial Analysis Coding Agent Index, 2.8 points above Fable 5. OpenAI said the model used less than half the output tokens, took less than half the time, and cost about one-third less than its competitor.

The company introduced a new capability setting called ultra, which coordinates four agents in parallel by default. This approach trades higher token use for stronger results and faster completion times on demanding tasks.

Itamar Friedman, co-founder and CEO at Qodo, said GPT-5.6 was the strongest model the company evaluated on its agentic code-review tests. He noted it beat GPT-5.5 on F1 while using roughly three times fewer tokens per pull request and delivering about two times lower median latency.

On knowledge work evaluations, GPT-5.6 Sol set new results on BrowseComp at 92.2% and OSWorld 2.0 at 62.6%. On OSWorld, it surpassed Opus 4.8 while using 85% fewer output tokens.

For cybersecurity applications, GPT-5.6 scored 73.5% on ExploitBench1 compared to GPT-5.5's 47.9% at a comparable output-token budget. On ExploitGym2, it reached 24.9% under a two-hour cap, nearly double GPT-5.5's peak pass rate of 15.1%.

OpenAI said GPT-5.6 models are more capable than earlier models in both biology and cybersecurity but do not cross the Critical threshold in either category. The company deployed layered safeguards that include protections trained into the model alongside real-time checks, continuous monitoring, and account-level enforcement.

The models underwent approximately 700,000 A100e GPU hours of black-box automated red teaming before launch, along with extensive red teaming by human experts.

GPT-5.6 is available starting Thursday across ChatGPT, Codex, and the OpenAI API. The rollout is beginning globally and will continue gradually toward full availability over the next 24 hours.

API pricing is set at $5 input and $30 output per 1 million tokens for Sol, $2.50 input and $15 output for Terra, and $1 input and $6 output for Luna. Cache writes are billed at 1.25 times the model's uncached input rate, while cache reads receive a 90% cached-input discount.

You May Also Be Interested In

• Stripe, Advent In Talks To Buy Paypal - WSJ
• Movano (MOVE) Misses Q2 EPS by 195c
• Home Depot (HD) call put ratio 1.5 calls to 1 put with a focus on September 330 puts into quarter results