Openai Unveils Cost-effective GPT-6 Models With Fewer Errors
WHO:
Techmeme - OpenAI is the organization launching the new models, GPT-6 Sol and GPT-6 Luna.
WHAT:
Techmeme - OpenAI's GPT-6 Sol and Luna models are described as budget-friendly versions of its most powerful model, with Sol priced at $2 per million input tokens and $10 per million output tokens, while Luna is priced at $0.10 per million input tokens and $0.50 per million output tokens.
Techmeme - OpenAI claims that GPT-6 Sol makes about half as many mistakes as GPT-5.6 Sol, and GPT-6 Luna matches GPT-5.6 Sol's performance at approximately 1% of the cost.
TechCrunch - OpenAI's new models are said to prioritize faster speeds and lower prices, with both models being about 50% cheaper than GPT-5.6.
Techmeme - OpenAI's GPT-6 Sol and Luna models are now available in the API, ChatGPT Work, and Codex.
Reddit ChatGPT Pro - GPT-6 Sol scored 68.8% at max effort on DeepSWE benchmarks, while Claude Fable 5 scored 69.9%, indicating a close performance at around 80% less cost per task.
Reddit ChatGPT Pro - GPT-6 Luna achieved a score of 66.6% on the same benchmark, comparable to Opus 5 and Fable 5 at medium effort, while costing 93 to 96% less.
IJR - OpenAI's GPT-6 Sol and Luna models were launched at half the prior API prices, with Sol priced at $2 per million input tokens and $10 per million output tokens, while Luna is priced at $0.10 per million input tokens and $0.50 per million output tokens.
Techmeme - Both models are about 50% cheaper than GPT-5.6.
Techmeme - GPT-6 Sol makes about half as many mistakes as GPT-5.6 Sol.
Techmeme - GPT-6 Luna matches GPT-5.6 Sol's performance at approximately 1% of the cost.
Techmeme - OpenAI's GPT-6 Sol and Luna models are described as being cut from the same cloth as Astra, indicating a shared technological foundation.
TechCrunch - OpenAI's new models are said to prioritize faster speeds and lower prices.
Mashable - GPT-6 Sol and Luna were launched with a lower propensity for mistakes compared to previous models.
Reddit ChatGPT Pro - OpenAI confirmed that the new prices for GPT-6 Sol and Luna are permanent, not promotional.
MarkTechPost - OpenAI's GPT-6 Sol and Luna models are now available in the API, ChatGPT Work, and Codex.
MarkTechPost - The new models come with improved prompt caching for long-running agents.
Reddit ChatGPT Pro - OpenAI confirmed that the new prices for GPT-6 Sol and Luna are permanent, not promotional.
Reddit ChatGPT Pro - GPT-6 Sol scored 68.8% at max effort on DeepSWE benchmarks, while Claude Fable 5 scored 69.9%, indicating a close performance at around 80% less cost per task.
Reddit ChatGPT Pro - GPT-6 Luna achieved a score of 66.6% on the same benchmark, comparable to Opus 5 and Fable 5 at medium effort, while costing 93 to 96% less.
Mashable - GPT-6 Sol and Luna were launched with a lower propensity for mistakes compared to previous models.
IJR - OpenAI's GPT-6 Sol and Luna models were launched at half the prior API prices, with Sol priced at $2 per million input tokens and $10 per million output tokens, while Luna is priced at $0.10 per million input tokens and $0.50 per million output tokens.
TechCrunch - OpenAI's new models prioritize faster speeds and lower prices.
Techmeme - OpenAI unveiled GPT-6 Sol and GPT-6 Luna, which are budget-friendly versions of its powerful model.
Techmeme - GPT-6 Sol costs $2 per 1 million input tokens and $10 per 1 million output tokens.
Techmeme - GPT-6 Luna costs $0.10 per 1 million input tokens and $0.50 per 1 million output tokens.
Techmeme - Both models are about 50% cheaper than GPT-5.6.
Techmeme - GPT-6 Sol makes about half as many mistakes as GPT-5.6 Sol.
Techmeme - GPT-6 Luna matches GPT-5.6 Sol's performance at approximately 1% of the cost.
WHEN:
IJR - OpenAI launched GPT-6 Sol and Luna on Sept. 22, 2026, approximately 90 minutes after Anthropic’s Opus 5.5.
IJR - OpenAI launched GPT-6 Sol and Luna on Sept. 22, 2026, approximately 90 minutes after Anthropic’s Opus 5.5.
Techmeme - GPT-6 Sol and Luna were launched less than three months after GPT-5.6.
WHY:
Techmeme - OpenAI claims major accuracy gains and dramatic cost reductions with the new models.
Techmeme - OpenAI claims major accuracy gains and dramatic cost reductions with the new models.
HOW:
MarkTechPost - The new models come with improved prompt caching for long-running agents.
Techmeme - The models are described as being cut from the same cloth as Astra, indicating a shared technological foundation.
(UPDATED September 23, 2026 13:58 UTC)
OpenAI has unveiled its latest models, GPT-6 Sol and Luna, which are designed to be budget-friendly alternatives to its most powerful offerings, with Sol priced at $2 per million input tokens and $10 per million output tokens, while Luna costs just $0.10 per million input tokens and $0.50 per million output tokens. The company claims that GPT-6 Sol makes about half as many mistakes as its predecessor, GPT-5.6 Sol, and that GPT-6 Luna performs comparably to GPT-5.6 Sol at roughly 1% of the cost. Additionally, these new models feature improved prompt caching for long-running agents, emphasizing faster speeds and lower prices, making them about 50% cheaper than GPT-5.6.
--
(UPDATED September 23, 2026 13:03 UTC)
OpenAI has just launched its GPT-6 Sol and Luna models, now available in the API, ChatGPT Work, and Codex, and both models are priced at about half of the previous API costs. In performance tests, GPT-6 Sol scored 68.8% on the DeepSWE benchmarks, closely trailing Claude Fable 5, while GPT-6 Luna achieved a score of 66.6%, comparable to Opus 5 and Fable 5, but at a fraction of the cost. Notably, GPT-6 Sol makes about half as many mistakes as its predecessor, GPT-5.6 Sol, and Luna matches its performance at roughly 1% of the cost.
--
(UPDATED September 23, 2026 10:58 UTC)
OpenAI has launched its new GPT-6 Sol and Luna models, which share a technological foundation with Astra and are designed to offer faster speeds and lower prices. These models also boast a lower propensity for mistakes compared to their predecessors, and OpenAI has confirmed that the new pricing is permanent rather than promotional. The launch took place on September 22, 2026, just 90 minutes after Anthropic’s Opus 5.5 was introduced.
--
(UPDATED September 23, 2026 05:59 UTC)
OpenAI has just released its GPT-6 Sol and Luna models, which are now accessible through the API, ChatGPT Work, and Codex. These new models feature enhanced prompt caching designed for long-running agents, improving their efficiency and performance. This update marks a significant advancement in the capabilities of OpenAI's technology.
--
(UPDATED September 23, 2026 02:04 UTC)
OpenAI has confirmed that the new prices for GPT-6 Sol and Luna are permanent, rather than promotional. In performance tests, GPT-6 Sol achieved a score of 68.8% at max effort on the DeepSWE benchmarks, closely trailing Claude Fable 5, which scored 69.9%, all while being around 80% less expensive per task. Meanwhile, GPT-6 Luna scored 66.6% on the same benchmarks, comparable to Opus 5 and Fable 5 at medium effort, but it costs an impressive 93 to 96% less.
--
(UPDATED September 22, 2026 23:57 UTC)
In a significant development in artificial intelligence, GPT-6 Sol and Luna have been launched, boasting a lower propensity for mistakes compared to their predecessors. This advancement aims to enhance user experience and reliability in various applications. With these new models, users can expect improved performance and accuracy in their interactions.
--
(UPDATED September 22, 2026 22:58 UTC)
OpenAI has just launched its new GPT-6 Sol and Luna models, offering them at half the previous API prices. The Sol model is priced at $2 per million input tokens and $10 per million output tokens, while the Luna model comes in at just $0.10 per million input tokens and $0.50 per million output tokens. This significant price reduction is expected to make advanced AI more accessible to developers and businesses alike.
--
(UPDATED September 22, 2026 22:01 UTC)
OpenAI has officially launched GPT-6 Sol and Luna, marking a significant advancement in artificial intelligence technology. This release took place at 22:01 UTC, just about 90 minutes after Anthropic introduced their own model, Opus 5.5. The competition in the AI landscape continues to heat up as these companies push the boundaries of what's possible.
--
(UPDATED September 22, 2026 20:00 UTC)
OpenAI has announced new models that focus on delivering faster speeds and lower prices. This development was reported by TechCrunch at 20:00 UTC. The changes aim to enhance accessibility and efficiency in AI technology.
--
(UPDATED September 22, 2026 19:01 UTC)
OpenAI has just launched two new models, GPT-6 Sol and GPT-6 Luna, which are designed to be budget-friendly alternatives to their powerful predecessors. GPT-6 Sol is priced at $2 per 1 million input tokens and $10 per 1 million output tokens, while GPT-6 Luna comes in even lower at $0.10 for input tokens and $0.50 for output tokens, making both models about 50% cheaper than GPT-5.6. OpenAI claims that these new models not only offer major accuracy gains but also maintain performance levels comparable to GPT-5.6 Sol, with Luna achieving this at approximately 1% of the cost.