GPT-6 Sol

OpenAI announced GPT-6 Sol and GPT-6 Luna as expansions of the GPT-6 generation following the earlier release of GPT-6 Astra, positioned to distribute frontier intelligence at lower cost. GPT-6 Sol is trained with similar methods as Astra, bringing advances in professional work, factuality, coding, computer use, and al The pricing table sets GPT-6 Sol at $2 per million input tokens and $10 per million output tokens, down from $4 and $20 respectively for GPT-5.6 Sol. On AutomationBench, GPT-6 Sol at xhigh effort is claimed to outperform Claude Opus 5 at max effort at just 9% of Opus 5's cost per task, while Luna is positioned for high

The llm-stats.com page presents benchmark and pricing data sourced from OpenAI's September 22, 2026 GPT-6 Sol launch. GPT-6 Sol ranks 21st on the LLM Stats Score composite with a score of 49.4, and has a blended price of $2.38 per million tokens. On DeepSWE 1.1 (software engineering agents using mini-swe-agent), GPT-6 On OSWorld 2.0, GPT-6 Sol ranks 6th with a score of 0.64/1 at max effort, at a cost of $3.25 per task. On Agents' Last Exam, it ranks 2nd with a score of 0.56/1. The page notes these evaluations used OpenAI research environments or API and may differ from production ChatGPT. (Note: the supplied excerpt is partially blo

OpenAI released GPT-6 Sol and GPT-6 Luna on September 22, 2026, completing the GPT-6 family under the flagship Astra that shipped on September 3. The pitch is not a smarter model than Astra but the same generation at roughly half the promotional token price, reshaping which model is worth reaching for in coding, research, and high-volume document work. No independent benchmark data existed at launch, and OpenAI published only selected comparisons that leave out Anthropic's Claude Opus 5.5, released the same evening. API pricing is $2 per million input tokens and $10 output for Sol and $0.10 and $0.50 for Luna, with cached input at 10% of the input rate and prompts over 272,000 tokens billed at long-context rates. Both share a 1,050,000-token context window and six reasoning-effort levels, ship in ChatGPT Work and Codex for Plus through Edu, and appear in the API as gpt-6-sol and gpt-6-luna, with Luna also in the desktop app for Free and Go. Sol targets complex coding and agentic workflows, Luna targets focused high-volume tasks like summarization and extraction, and neither is in the Chat surface yet, while GPT-5.5 retires from ChatGPT, ChatGPT Work, and Codex on October 14, 2026.

The supplied article text consists almost entirely of cookie-consent notices, site navigation menus, and unrelated headline snippets from The New Stack, with no substantive content about OpenAI releasing GPT-6 Sol and Luna or cutting token prices in half. The headline and metadata alone indicate OpenAI announced two new models, Sol and Luna, alongside a halving of token pricing. No details, quotes, benchmarks, or release dates appear in the extractable body text. Because the firecrawl extraction returned only boilerplate, paragraphs must be limited to what is explicitly stated in the title and source metadata: the article was published on The New Stack on September 22, 2026, and reports on OpenAI's release of GPT-6 Sol and Luna together with a fifty-percent reduction in token prices. Any further specifics about capabilities, pricing tiers, availability, or context cannot be confirmed from the supplied text and are therefore omitted.

OpenAI launched GPT-6 Sol and Luna as updated versions of the smaller Sol and Luna models within the GPT-6 generation, following the earlier launch of GPT-6 Astra. The TechCrunch article reports that OpenAI describes the new Sol and Luna as extending the GPT-6 generation's benefits by making its intelligence more effic According to the supplied excerpt, the GPT-6 series models are available at half the cost of the 5.6 series of Sol and Luna, a price drop OpenAI attributes to improvements in caching and inference. OpenAI claims the newest models produce more factually accurate results and have a lower error rate on coding. Specificall