Vibeleaderboard
← All Intel
Intel / post

LangSmith Signal: gpt-6-luna Becomes Most-Used Production Model

Source
x.com
Date
LangChain@LangChain

gpt-6-luna launched Sept 22. Today, more production teams are using it than any other model. The latest LangSmith Signal ⤵️ ✅ gpt-6-luna: #30 → #1. Used by 27% of production teams. ✅ claude-sonnet-5-5 + gpt-6.1-sol entered at #5 + #6 on our adoption list at 22% each. Each model handles <2% of LLM calls. ✅ gpt-5.6-luna initially handled the most LLM calls, but was surpassed on a weekly basis by gpt-6-luna as of Nov 2.

Context

LangChain, whose LangSmith service records the model calls that companies' AI applications make, says gpt-6-luna is now used by more production teams than any other model in its data (post). OpenAI released gpt-6-luna, its low-cost model, on September 22 at half the input price of gpt-5.6-luna and less than half its output price, according to Simon Willison's launch write-up. In LangSmith's ranking for September 28 to November 14, 27% of production teams used it, and it rose from 30th place to first. LangChain counts as production teams the organizations on paid plans or with at least 17,500 calls in that period, and the figures come from LangSmith's own API traffic.

The two charts attached to the post measure different things, and the gap between them is the useful part. The adoption chart counts the share of teams that used a model at all: gpt-6-luna leads at 27%, followed by gpt-5.6-luna and Anthropic's claude-haiku-4-5 at 24% each. The volume chart counts the share of all LLM calls, and over the same weeks gpt-5.6-luna still carried the most, 13% against 9% for gpt-6-luna, although LangChain says gpt-6-luna passed it on a weekly basis from November 2. The sharpest gap is for two other new models, claude-sonnet-5-5 and gpt-6.1-sol, which entered the adoption list at fifth and sixth with 22% of teams each yet handle less than 2% of calls each. Many teams have connected them, but little traffic runs through them. The post does not say why, and the data cannot show whether those teams are testing the models or using them for a few demanding tasks.

These are LangChain's numbers about its own customers, not the whole market. The post and its charts give no count of teams or calls, and they do not state the period behind the jump from 30th place. If it is August 11 to September 27, the earlier window the volume chart uses for growth, gpt-6-luna was available for less than a week of it, which would account for much of the climb. Within those limits the result is clear: among LangSmith's production teams, gpt-6-luna reached the widest use within eight weeks of launch, while its predecessor, gpt-5.6-luna, carried more calls than it until early November. An earlier LangSmith Signal tracked the rise of models in LangSmith's data.

Terms in this piece · Glossary
  • LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
  • open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
More from LangChain
Recommended reads
Comments

Checking sign-in…

Loading comments…