OpenAI launches GPT-6.1 Sol: near-Astra performance at one-fifth the price
GPT-6.1 Sol is an upgrade to GPT-6 Sol that nearly matches GPT-6 Astra on coding, computer use, and professional work benchmarks. Standard input is $2, cached input $0.10, standard output $10 per million tokens.
OpenAI launched GPT-6.1 Sol on September 29, positioned as the mid-tier workhorse that nearly matches GPT-6 Astra intelligence at one-fifth the price. API pricing is $2 standard input, $0.10 cached input, and $10 standard output per million tokens. Cached input is 95% cheaper than standard input and 50% cheaper than GPT-6 Sol cached input. By comparison, GPT-6 Astra is $10 input and $50 output; GPT-6 Luna is $0.10 input and $0.50 output per million tokens.
On the DeepSWE v1.1 coding benchmark, GPT-6.1 Sol matches GPT-6 Astra at about one-fifth the cost and beats GPT-6 Sol's best by 6.4 percentage points at lower inference intensity. On the GDP.pdf professional document Q&A benchmark, GPT-6.1 Sol beats Opus 5.5 at less than half the cost and approaches GPT-6 Astra's state-of-the-art at roughly one-fifth the cost. On the AutomationBench multi-step business workflow benchmark, GPT-6.1 Sol at medium inference beats Opus 5.5 by 2.2 percentage points at about one-third the cost and beats GPT-6 Sol by 4.8 percentage points.
On the OSWorld 2.0 offline computer use benchmark at maximum inference intensity, GPT-6.1 Sol beats GPT-6 Sol by 7 percentage points at less than half the cost, with only a 2.1 percentage point gap to Astra at roughly one-seventh the cost. On Terminal-Bench Science 0.1 at maximum intensity, GPT-6.1 Sol more than doubles GPT-6 Sol's score at $5.47 average per task (versus Opus 5.5 at $23.21 and Astra at $23.80), over 75% cheaper than both. Astra remains the highest scoring model at 68.1%, so the most difficult scientific research tasks should still use Astra.
On factuality, GPT-6.1 Sol reduces responses containing factual errors at low inference intensity from 11.4% to 7.7%, a roughly 32% reduction, with error rate within 1.9 percentage points of Astra. In the broken-search-tool alignment evaluation, undisclosed failure rate is only 2.1% (versus GPT-6 Sol at 4.9%), on par with Astra. Full safety details are available in the GPT-6.1 Sol system card addendum at deploymentsafety.openai.com/gpt-6-1-sol.
GPT-6.1 Sol is available immediately via the API (model name gpt-6.1-sol) and to Plus, Pro, Business, Enterprise, and Edu users, though it is not yet available in Chat conversations. GPT-6.1 Sol Ultrafast will launch in the coming days and run up to 8x faster than standard in Codex.