OpenAI Cuts GPT-5.6 Luna Pricing by About 80%

The number of parameters actually executed during each inference forward pass, expressed in billions. For Mixture of Experts (MoE) models, a routing mechanism selects a subset of experts per token, resulting in fewer active than total parameters. Dense models use all parameters, so active equals total. ## Frequently Asked Questions GPT-5.4 (xhigh) is more intelligent. GPT-5.4 (xhigh) scores 51, compared with GPT-5.6 Luna (high) at 46 on the Artificial Analysis Intelligence Index. GPT-5.6 Luna (high) is faster. GPT-5.6 Luna (high) generates 166.2 tokens per second, compared with GPT-5.4 (xhigh) at 127.0 tokens per second. [...] GPT-5.6 Luna (high) is cheaper. GPT-5.6 Luna (high) costs $0.17 per 1M tokens, compared with GPT-5.4 (xhigh) at $2.17 per 1M tokens (7:2:1 cache hit/input/output ratio). GPT-5.6 Luna (high) has lower latency. GPT-5.6 Luna (high) has a time to first token of 8.70s, compared with GPT-5.4 (xhigh) at 142.98s. GPT-5.4 (xhigh) has a larger context window. GPT-5.4

The curve is clear. Luna moves from 26.6 without reasoning to 51.2 at max, a gain of about 24.7 index points. The full-index evaluation cost rises by roughly 13.5 times, and weighted output tokens rise by roughly 8.9 times. The price per token did not change between these rows. The effective cost changed because the model used more tokens and more time. This is what I mean by the price of thinking. A reasoning setting is not simply a quality toggle. It is a compute allocation. For an easy task, max may be wasteful. For a difficult task, no reasoning may be cheap but unreliable. The economic decision is not “Which model is best?” It is “Which amount of test-time computation produces a verified result at an acceptable cost and delay?” [...] OpenAI’s GPT 5.6 Luna has become a much more consequential model overnight. On July 30, OpenAI announced an 80% price reduction for Luna and a 20% reduction for Terra. Luna now costs $0.20 per million input tokens and $1.20 per million output tokens.

feel perfectly content in using pay as you go pricing with deepseek. On the other hand, although Anthropic's models used to be my bread and butter for personal work, they are simply too expensive to reach for these days. reply | | | | | | rpdillon 4 days ago | root | parent | prev | next (javascript:void(0)) This is exactly the model that DeepSeek V4 Flash followed, and it's been insanely successful as a result, even though it's not frontier. reply | | | | | | ismailmaj 4 days ago | parent | prev | next (javascript:void(0)) it's 80% less cost, not 80% in efficiency gains, could be that Luna was overpriced to begin with, we don't have much info on the models themselves. Assuming the efficiency gains are real, I feel like something has to give, maybe worse quality due to [...] $0.37/task - Luna medium: score 38, approximately $0.01/task - Luna max: score 51, approximately $0.042/task So Luna medium is: - slightly more capable than GPT-5 high; - approximately 35–40× cheap

Starting today, GPT‑5.6 Luna, our fastest and most affordable model, will cost 80% less, while GPT‑5.6 Terra, our balanced model for everyday work, will cost 20% less. These lower prices for Luna and Terra are also reflected in how usage is counted against paid subscriptions when using Codex and ChatGPT Work. Luna gives businesses a far more cost-effective way to handle high-volume work at very high levels of quality. It can use tools and complete multi-step workflows, making a broader range of AI applications practical to run at scale. [...] Hoda Noorian, AI Product, Notion > “We are focused on maximizing outcome per dollar in everything we do, and the GPT‑5.6 series has been a significant step forward in that regard. Terra and Luna lead on cost-efficiency across our internal coding benchmarks, and Luna is now our default model for background agent automations.” Shaiyon Hariri, Creator of Ramp SWE-Bench > “GPT‑5.6 Luna is the biggest step change in agentic behavior we’ve seen since

The big picture:OpenAI claims it can now deliver substantially more intelligence per dollar through efficiency improvements. Cheaper Chinese open-weight models have also increased pressure on OpenAI and Anthropic to prove that their models justify their higher costs. Driving the news:OpenAI will lower the price of GPT-5.6 Luna — the fastest and lowest cost version of the new model— by roughly 80%. Luna will now cost 20 cents per million input tokens and $1.20 per million output tokens. OpenAI says it will reduce the price of GPT-5.6 Terra — the mid-range version — by 20%. Use of that version will now cost $2 per million input tokens and $12 per million output tokens. The most powerful version — GPT-5.6 Sol — won't see a price cut. [...] Consent powered by Ethyca Skip to main content sms (opens in new window) facebook (opens in new window) twitter (opens in new window) linkedin (opens in new window) bluesky (opens in new window) Add Axios on Google Add Axios

Notably, GPT-5.6 Soul’s own price didn’t drop, even though it almost certainly saw similar efficiency gains internally. The likely explanation is that Soul is where the bulk of OpenAI’s revenue comes from, not the smaller workhorse models, so any efficiency gains there get absorbed as margin rather than passed to customers. Competitive pressure, particularly from open-source releases like Kimi K3 out of China, may be part of what pushed OpenAI to move quickly on Luna and Terra’s pricing instead. ## Frequently Asked Questions ### Why did GPT-5.6 Luna get an 80% price cut but Soul got none? [...] ## What did OpenAI actually announce? OpenAI cut the price of GPT-5.6 Luna, the smallest model in its GPT-5.6 family, by 80%. It now costs 20 cents per million input tokens and $1.20 per million output tokens, undercutting most frontier open-source models coming out of China on a per-token basis. GPT-5.6 Terra, the mid-tier model, got a 20% price cut, landing at $2 per million input tokens an