GPT-6 Luna
One of the models in OpenAI's GPT-6 series, released alongside Sol. Built on the same Astra-derived architecture and positioned for low cost and low inference error rates, it may be oriented toward lightweight, high-throughput use cases.
Core Facts
The API model identifier for GPT-6 Astra is gpt-6-astra.
GPT-6 Sol and GPT-6 Luna deliver strong performance at lower cost and have pushed the Pareto frontier on the OfficeQA Pro benchmark
GPT-5.6 Sol is an all-new flagship model focused on complex reasoning and agentic programming.
Timeline (last 90 days)
Free users' GPT-6 Luna has relatively strict rate limits and may be downgraded or queued during peak periods; Go users get a more stable access priority and a higher message cap.
After GPT-6 lands, ChatGPT free users and the Go tier ($8/month) can both call GPT-6 Luna in Chat conversations, with a very small gap in experience between the two, and Go's added value mainly shows up in a higher usage quota rather than in the model itself.
Luna (Latin for "moon") and Sol (Latin for "sun") are the sub-model suffixes OpenAI enabled in the GPT-6 generation, using celestial metaphors to distinguish different compute configurations of the same generation of models.
GPT-6 Luna's pricing tier is set at 272,000 tokens, which kicks in later than Haiku's 100,000-token threshold.
In LeanZero's two agentic coding benchmarks, Haiku 5.5 won both rounds, with a combined billing of $1.62 for the two runs versus Luna's $0.46, a difference of approximately 3.6 times.
On the Forge 1.0 benchmark, Haiku scored 0.9661 and Luna scored 0.395, with Haiku trailing Claude Opus 5.5 by only 0.0106.
On the Gauntlet 7.2 benchmark, Haiku scored 0.661 and Luna scored 0.4797.
Luna never imported historical changes during the Forge test—none of the 62 historical changes were included—triggering a backfill check labeled as critical, which multiplied the entire round's score by 0.6, changing it from 0.6583 to 0.395.
Luna's manifest was missing the scope required for Jira Software calls (causing 401 responses from the site), and it called @forge/llm without declaring the llm module (causing the interpretation feature to score 0).
The author argues that the real driver of the cost gap is not that Haiku produces more output, but rather its earlier-arriving pricing tier; without the tier, Gauntlet would cost approximately $0.30 and Forge approximately $0.19, roughly on par with Luna's $0.23 and $0.22.
40 more timeline events
All Facts (20)
The API model identifier for GPT-6 Astra is gpt-6-astra.
90%VerifiedGPT-6 Sol and GPT-6 Luna deliver strong performance at lower cost and have pushed the Pareto frontier on the OfficeQA Pro benchmark
75%VerifiedGPT-5.6 Sol is an all-new flagship model focused on complex reasoning and agentic programming.
75%VerifiedThe Decisions API runs on a model called GPT-6 Luna
70%VerifiedGPT-4.1 Sol is now available in the ChatGPT Workbench and Codex for users from Plus to EDU tiers, but has not yet launched in the regular chat interface.
70%VerifiedOpenAI officially stated that GPT-6 Sol and Luna share the same lineage as the previously released Astra series.
70%VerifiedGPT-6 Luna can handle almost all tasks previously done with GPT-5.6 Soul, including making PPTs, video editing with subtitles, creating covers, automated publishing workflows, complex Skill modifications, simple project development, and researching new technologies.
65%VerifiedGPT-6 Astra is an entirely new pre-trained model generation, not a minor version iteration.
65%VerifiedGPT-6 Sol and Anthropic's Opus 5.5 were released on the same day.
70%VerifiedDevin officially announced that the newly integrated GPT-5.6 Sol model is now live on both the Devin Desktop and Devin CLI platforms
65%VerifiedGPT-5.6 Luna has been made available to free users of ChatGPT.
65%UnverifiedThe new model simultaneously achieves higher quality and lower cost, breaking the traditional performance-cost trade-off.
70%Partially VerifiedThe test covered four vision models: Claude Sonnet 5 (Anthropic), Gemini 3 Flash (Google), Gemini 3.1 Flash-Lite (Google), and GPT-5.4 nano (OpenAI).
65%Partially VerifiedCompared to the previous generation, Sol's input and output prices both dropped by half, while Luna's input price dropped by half and its output price dropped by approximately 58%.
65%Partially VerifiedGPT-6 Sol's price dropped by half while its intelligence level remained almost on par with the previous generation.
65%Partially VerifiedOpenAI announced that the Decisions API is open for public testing to all developers, powered by GPT-6 Luna and supporting text and image input.
65%UnverifiedFree users' GPT-6 Luna has relatively strict rate limits and may be downgraded or queued during peak periods; Go users get a more stable access priority and a higher message cap.
50%UnverifiedLuna (Latin for "moon") and Sol (Latin for "sun") are the sub-model suffixes OpenAI enabled in the GPT-6 generation, using celestial metaphors to distinguish different compute configurations of the same generation of models.
50%UnverifiedAfter GPT-6 lands, ChatGPT free users and the Go tier ($8/month) can both call GPT-6 Luna in Chat conversations, with a very small gap in experience between the two, and Go's added value mainly shows up in a higher usage quota rather than in the model itself.
50%UnverifiedOn the Forge 1.0 benchmark, Haiku scored 0.9661 and Luna scored 0.395, with Haiku trailing Claude Opus 5.5 by only 0.0106.
50%