OpenAI: GPT-6 Luna
Kept by incident.models@web-summit.dinis, 11 November 2026, 03:06 UTC
Not a warning service: for warnings follow IPMA and ANEPC. Nothing here is a forecast or advice. One of the models in Other models, measured. The id we call it by: openai/gpt-6-luna (through OpenRouter).
What it is
- Publisher: OpenAI.
- Open weights: no. It is used only through its provider's API (here by way of OpenRouter).
- Licence: OpenAI's commercial terms for its API; no weights are published.
- Reasoning:
openai/gpt-6-luna: off (the provider's default);openai/gpt-6-luna#reasoning-on: on (asked, medium effort). - List price (OpenRouter, read 2026-11-11 02:19 UTC): 0.1 US dollars per million tokens in, 0.5 out. Listed there since 2026-09-22. Our grouping: commercial, small.
Its use here, in one line
89 runs on 5 tasks, 82 right; 0 not answered (malformed, cut off, refused or a provider error); 73,120 tokens in, 8,880 out; 0.00958 US dollars. Source: results.json, made 2026-11-11 03:06 UTC. It has written nothing on this site: every call was a test.
Every job we gave it
| Task | Setting | Runs right | Items right, run by run | Malformed | Cut off | Refused | Errors | Tokens in | Tokens out | Cost, USD | Per item, USD | Median s | p90 s |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Warnings in force at an hour | off | 15/17 | 8 of 9 (1 run) | 0 | 0 | 0 | 0 | 42551 | 1990 | 0.00312 | 0.00018 | 2.9 | 4.0 |
| Warnings in force at an hour | on | 5/5 | 5 of 5 (1 run) | 0 | 0 | 0 | 0 | 12515 | 763 | 0.00164 | 0.00033 | 3.4 | 4.3 |
| Extract a decision | off | 9/10 | 9 of 10 (1 run) | 0 | 0 | 0 | 0 | 6694 | 1177 | 0.00125 | 0.00012 | 2.6 | 4.9 |
| Label fact, evidence, hypothesis | off | 16/20 | 16 of 20 (1 run) | 0 | 0 | 0 | 0 | 5486 | 1904 | 0.00149 | 0.000074 | 2.1 | 4.0 |
| Translate, figures kept | off | 20/20 | 20 of 20 (1 run) | 0 | 0 | 0 | 0 | 1701 | 1689 | 0.00100 | 0.000050 | 2.9 | 3.9 |
| Extract a warning | off | 17/17 | 17 of 17 (1 run) | 0 | 0 | 0 | 0 | 4173 | 1357 | 0.00109 | 0.000064 | 2.6 | 2.8 |
What we would use it for
Ours, a hypothesis from these runs. A cheap commercial second reader: 20 of 20 translations, 16 of 20 labels, 9 of 10 decisions (it named the Government as the decider of Web Summit's own decision), and 8 of the 9 harder warnings items it reached. About the price of the open-weight models. No reason to prefer it to them here, and no reason to avoid it as the second opinion from a different maker.
The items it got wrong
First run of each setting; the item links to its page, where every model's answer is.
Warnings in force at an hour, off (1 of 9):
| Item | Scored against | It answered | Result |
|---|---|---|---|
| f01 | [{"type":"coastal_event","level":"yellow","end_utc":"2026-11-11T06:00:00Z"},{"type":"rain","level":"orange","end_utc":"2026-11-11T09:00:00Z"}] | [{"type":"rain","level":"orange","end_utc":"2026-11-11T09:00:00Z"}] | wrong |
Extract a decision, off (1 of 10):
| Item | Scored against | It answered | Result |
|---|---|---|---|
| d07 | {"decider":"websummit","effect_from_utc":null,"effect_to_utc":null} | {"decider":"government","what":"Day 2 of Web Summit (Wednesday, November 11) will not take place.","effect_from_utc":null,"effect_to_utc":null} | wrong |
Label fact, evidence, hypothesis, off (4 of 20):
| Item | Scored against | It answered | Result |
|---|---|---|---|
| s-1110T2115 | evidence | evidence | wrong |
| s-1110T2153 | fact | hypothesis | wrong |
| s-1110T2358 | hypothesis | fact | wrong |
| s-1111T0015 | hypothesis | fact | wrong |