GPT Luna
OpenAI's small, low-cost GPT model tier (GPT-5.6 / GPT-6 Luna) for API and ChatGPT/Codex use.
Liked as a very cheap workhorse model1234 for routine coding, subagents and high-volume batch tasks.5678 The most common complaint is that it reasons less well than larger models and misses obvious things,9101112 with some saying its quality has slipped and that high-reasoning runs are slow.131415
Why people recommend it
Cost efficiency · 3
Commenters repeatedly call it extremely cheap for what it delivers, and some moved whole workflows to it to cut costs.
“Luna and Sol-6.1 are crazy cost effective and all flows here have switched to them”
“luna is so unbelievable cheap.”
Show 1 more
“Luna is insanely good for the cost and it's my go-to for high volume batch tasks now.”
Routine workhorse · 3
Used as the default model for everyday implementation, simple edits and delegated subagent work, while bigger models handle planning.
“To me Luna is the work horse”
“in codex - luna high for 80% of everything.”
Show 1 more
“Luna is great at doing targeted smaller work”
Speed · 3
Described as small and fast, which suits interactive, hands-on work.
“Luna is small and fast”
“i enjoy it a lot more if i use a super fast model with low reasoning (gpt 6 luna low effort + fast mode)”
Show 1 more
“I found GPT 5.6 LUNA to be astonishingly capable.”
Stretches plan limits · 3
Users say it uses little quota on Plus and Codex plans, so they rarely hit usage limits.
“luna (max reasoning). price, as it sips quota on the Plus plan and is pretty competent for my needs.”
Show 1 more
“For personal use, Luna-xhigh on the Plus subscription.”
Batch processing · 3
Picked for high-volume jobs such as classifying large datasets and support-ticket dumps.
“Luna is insanely good for the cost and it's my go-to for high volume batch tasks now.”
“just paying the cheap Luna prices to openai and getting it done in a few hours”
Show 1 more
Code review · 2
Some find it good at focused code review when the task is narrowly scoped and split across reviewers.
“I found Luna and even 5.4-mini to be quite good at code review”
“six subagents (I recommend Luna)”
Caveats and dissent
Limited intelligence · 3
Several commenters say it scores lower on intelligence, misses obvious points and is unreliable for harder work.
“Luna is cost competitive, but doesn't score as well on intelligence.”
“Luna misses the obvious (and more importantly, the clearly explained) consistently.”
Show 1 more
Quality regression · 2
Some report the model seemed to get dumber over time and users are upset.
“over the last 2 or 3 weeks I've seen how dumb it is now”
“I've never seen users as upset about GPT-6 Sol/Luna than I have right now.”
Slow at high effort · 3
At xhigh reasoning it takes longer to finish tasks, and one user wants more tokens per second.
“Luna xhigh is more than good enough for me”
“Luna xhigh has been really, really good for the price.”
Show 1 more
“I love luna but unfortunately it needs to be much, much faster in tokens per second”
Needs tight scoping · 2
It works better when told exactly what to do or given a narrow focus.
“over the last 2 or 3 weeks I've seen how dumb it is now”
“six subagents (I recommend Luna)”
Often compared with
2 moved from GPT Sol to GPT Luna
3 moved from Claude Opus to GPT Luna
2 moved from DeepSeek V4 Flash to GPT Luna
1 moved from GPT Terra to GPT Luna
1 moved from Claude Haiku to GPT Luna
1 moved from GPT Luna to Claude Haiku
Who recommends it in their own words
Threads where it came up
- Ask HN: What default model do you use and why? 11 mentions
- Advancing the price-performance frontier with GPT‑5.6 8 mentions
- GPT-6 Sol and Luna 6 mentions
- GPT-5.6 Luna vs. GPT-6 Astra: Is a $1.20 Model Good Enough for Code Review? 6 mentions
- Claude Haiku 5.5 5 mentions
- Why isn't the industry freaking out about DeepSeek 4.1 Flash? 4 mentions
Also written as GPT-5.6 Luna, GPT 5.6 Luna, GPT-6 Luna, GPT 6 Luna. Data: gpt-luna.json. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.
All comments 50
“switched to Luna for less intelligent models”
“you can use software like Luna or Duet, or native screen sharing to control an iMac from another Mac”
“Luna and Sol-6.1 are crazy cost effective and all flows here have switched to them”
“Also should work in as pair of Astra + Luna in CodeX.”
“Luna is small and fast”
“On the API we use a lot of Luna (cost efficient)”
“Gemini Flash 3.8, DeepSeek Flash, and Luna are still up there as favorites”
“Luna seems more willing than Sol”
“Coding is handled by Haiku 5.5/GPT 6 Luna/DeepSeek v4.1 Flash.”
“Claude Haiku 5.5 and GPT-6 Luna - both excellent inexpensive models”
“My daily driver was Luna, with occasional elevation to Terra or Sol”
“We use OpenAI Luna for the agent37/default model”
“I love luna but unfortunately it needs to be much, much faster in tokens per second”
“Even cheap models like GPT 6 Luna do decent job in programming these days.”
“delegate large portion of them to gpt-luna”
“I often push down to Terra or Luna”
“my personal experience is Luna is better for what I do, and Luna is cheaper.”
“Luna is cost competitive, but doesn't score as well on intelligence.”
“Luna misses the obvious (and more importantly, the clearly explained) consistently.”
“luna is so unbelievable cheap.”
“Luna-max for planned coding and general tasks.”
“I use Luna for this day in and out and its excellent”
“used <10% of my 5hr limit on a $100 codex plan.”
“I use Luna for certain tasks where speed is more valuable than performance”
“Lunas output is not really helpful here.”
“i would be shocked if luna decides is less generally capable than jev”
curl https://api.openai.com/v1/decisions \ -H "Authorization: Bearer $(llm keys get openai)" \ -H "Content-Type: application/json" \ --data ' { "model": "gpt-6-luna", "input": [{ "role": "user", "content": [ {"type": "input_text", "text": "I am angry about the new product feature"} ] }], "questions": [{ "type": "predicate", "name": "complaint", "instructions": "Is this a complaint?" }, { "type": "predicate", "name":
“Luna as it was cheap and more than powerful enough”
“cheaper models like the GPT Luna family”
“I think Luna is a fantastic driver for these”
“GPT 6 Luna closed the gap significantly for sure”
“Don't get me started on Luna.”
“luna implemented the project”
“I've never seen users as upset about GPT-6 Sol/Luna than I have right now.”
“for the rest of the stuff I use sonnet 5 / gpt luna”
“I use gpt-6-sol for planing and luna for writing code”
“i enjoy it a lot more if i use a super fast model with low reasoning (gpt 6 luna low effort + fast mode)”
“luna for day to day tasks”
“in codex - luna high for 80% of everything.”
“Luna doesn't feel more capable nor faster than Qwen3.8”
“If it is something super simple like changing a css class I use Luna.”
“Luna xhigh is more than good enough for me”
“over the last 2 or 3 weeks I've seen how dumb it is now”
“Terra, Luna - for implementation in simple routine cases.”
“I use Luna for research/scout/test subagents”
“To me Luna is the work horse”
“I know Luna isn't reliable”
“Luna is the undisputed king here at the moment and is what I use as my workhorse model.”
“Luna is insanely good for the cost and it's my go-to for high volume batch tasks now.”
“just paying the cheap Luna prices to openai and getting it done in a few hours”